Monitoring Critical Infrastructure Using 3D LiDAR Point Clouds

Monitoring critical infrastructure is of great importance for resilience against current and emerging hazards. In Canada, at the national level, critical infrastructure is strategically classified into the following 10 sectors: energy and utilities, finance, food, transportation, government, information and communication technology, health, water, safety, and manufacturing. Many of these critical infrastructures may be effectively monitored using emerging technologies such as light detection and ranging (LiDAR). LiDAR technology enables active remote sensing from airborne and terrestrial systems. Accordingly, LiDAR can be used in a wide range of applications related to the monitoring of critical infrastructure to promote resilience. This survey provides a comprehensive, and structured literature overview of LiDAR technology currently available for commercial and research use, and also provides a detailed review of the relevant applications of LiDAR for critical infrastructure monitoring. Several LiDAR-based applications in areas of critical infrastructure such as terrestrial and air transportation, gas, oil, and water pipelines, energy generation and distribution, and security are discussed. A summary of the LiDAR data sets that are currently available for use is also presented.


I. INTRODUCTION
In Canada, critical infrastructure is strategically classified into several sectors, namely energy and utilities, finance, food, transportation, government, information and communication technology, health, water, safety, and manufacturing [1]. Active and continuous monitoring of critical infrastructure is required to promote resilience to a broad range of potential hazards. A range of sensing technologies, such as cameras, radar, satellite imagery, and seismographs, are available for such monitoring systems. LiDAR is a safe, relatively inexpensive emerging sensing technology that provides a three-dimensional (3D) representation of the environment, from which actionable conclusions may be extracted via data analysis and machine learning (ML) techniques.
The associate editor coordinating the review of this manuscript and approving it for publication was Okyay Kaynak .
LiDAR can be used in a wide range of applications related to the monitoring of critical infrastructure to promote resilience. Comprehensive survey on different types of LiDAR [2], [3], [4] have been extensively studied by the researchers. Individual surveys on different applications of LiDAR in road transportation [5], [6], [7], [8], [9], air transportation [10], [11], [12], [13], railway [14], [15], electrical power transmission [16] and people tracking [17] are present in literature. Several LiDAR databases are also available for post processing applications [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47]. However, a comprehensive survey that includes all different applications of LiDAR for infrastructure monitoring along with databases is not present in literature. The motivation behind this work was to highlight the different fields of application of LiDAR for critical infrastructure monitoring in a single survey to facilitate the advancement of research in LiDAR technology. All the above mentioned surveys incorporate the recent trends in the application of LiDAR technology for infrastructure monitoring using ML techniques. The most commonly used ML techniques are support vector machines (SVM), convolutional neural networks (CNN), principle component analysis (PCA) and k-means. Table 1. shows some survey papers and their applicability along with limitations of each.
The paper is organized as follows. In Section 2, the LiDAR technology, including its strengths and limitations, is explained. Section 2 also includes a discussion of publicly available LiDAR data sets for critical infrastructure monitoring. In Section 3, various applications of LiDAR for monitoring critical infrastructure, including transportation (road, rail, and air), pipelines (oil, gas, water), electrical power transmission and distribution equipment, mining, and security, are discussed. The paper concludes with recommendations for incorporating LiDAR for enhancements to the resilience of critical infrastructure in Canada in Section 4.

II. LiDAR TECHNOLOGY
This section provides background information for LiDAR technology. First, we briefly explain the different types of LiDAR and their most common specifications. Next, we explore the publicly available LiDAR data sets.
There are two types of optical remote sensing systems: passive and active. While passive remote sensing uses solar energy to measure the scattered light, active systems transmit signals and measure the scattered signal from objects [20]. Since the signals transmitted in active remote sensing are known, the measurement of the received signals is more controlled, which results in reduced artifacts. LiDAR is a form of active remote sensing.
LiDAR is a technology that employs light to measure the distance (range) from the measuring device to targets [20] and uses time of flight (ToF) to detect targets. There are two main techniques employed by LiDAR systems: pulsed and continuous-wave laser. Pulsed LiDAR sends small bursts of energy to the target to measure the intensity (amplitude) of the reflected signals. The continuous-wave laser transmits a modulated signal and measures the phase difference between transmitted and received signals [20]. The time difference between the transmitted and received signal enables the computation of the distance from the LiDAR device to the target obstacles. LiDAR active remote sensing can be used on airborne, spaceborne, and terrestrial applications [20].

A. TYPES OF LiDAR
There are different types of LiDAR technologies. The two available LiDAR types based on the platform are terrestrial and airborne. Most of the LiDAR systems commercially available are composed of rotating parts, where the laser beam scans 360 • around the LiDAR axis. ToF cameras, on the other hand, usually do not have a rotating part and can measure multiple points without changing the position of the sensor. They also measure shorter distances than LiDAR systems with a laser beam. In this section, different types of LiDAR technologies and methods of data collection are explored.

1) AIRBORNE
For airborne LiDAR, the data collection system is installed on an airplane, a helicopter or a drone. The light is transmitted from the airborne system to the ground, and the scattered light is detected by the system while in motion [21]. There are two main types of airborne sensors: the topographic and the bathymetric LiDAR.

a: TOPOGRAPHIC LiDAR
Topographic LiDAR collects surface cloud points that represent the elevation of the Earth's surface. This type of LiDAR is used in forestry, hydrology, geomorphology, urban planning, landscape ecology, coastal engineering, survey assessments, and volumetric calculations [21]. A topographic LiDAR sensor is given in Fig. 1.

b: BATHYMETRIC LiDAR
The bathymetric LiDAR is used to simultaneously measure the distance from the airborne system to the top layer of the water, as well as the ocean/sea's depths. Infrared laser is commonly used to measure the topography, and in the case of bathymetric LiDAR, an extra green laser is used to measure the bottom of ocean/sea bodies, providing two different measures. These measurements can then be used to extract the ocean/sea's depths and shoreline elevations [21]. This type of LiDAR is typically used for deployment near coastlines, shores and river banks or protecting harbours. It can also be used to locate objects on the ocean floor. Airborne LiDAR bathymetry (ALB) can be used to monitor the morphology of rivers with high degrees of accuracy [23]. Based on LiDAR applications such as detecting coral reefs, one can conclude that LiDAR can be used for monitoring water bodies for effects of climate change. Fig. 2 shows a bathymetric LiDAR sensor.

2) TERRESTRIAL LiDAR
Terrestrial LiDAR can be divided into two main classes: mobile and static. Mobile terrestrial LiDAR is mounted on a moving platform such as a vehicle, train or a boat while static terrestrial LiDAR is mounted on a tripod or a stationary device. Both types of LiDAR employ eye-safe lasers. These radars are used to conduct surveys on rails and highways, manage facilities, or create highly accurate 3D models for interior and exterior spaces [21].

a: MOBILE
The system usually consists of the LiDAR sensor along with additional sensors such as global positioning systems (GPS), inertial navigation systems (INS), and cameras. This type of system is used to obtain the location of encroaching overhead wires, high poles, road signs near roadways or rail lines [21]. Fig. 3 provides an example of mobile terrestrial LiDAR that was used to obtain details about the shore line, the point cloud view, the view after filtering the data and the cross-shore profile view. FIGURE 3. Mobile terrestrial LiDAR system [25] used for obtaining profile of a shore. The figure also contains the original data, filtered data and processed shore profile.

b: STATIC
Static LiDAR can be used to acquire data points from interior of buildings as well as external structures. This makes static LiDAR useful in areas such as engineering, mining, surveying, and archaeology [21]. Fig. 4 shows an example of static terrestrial LiDAR sensor.

B. HARDWARE AND SOFTWARE REQUIREMENTS
Most commercially available LiDAR sensors are configured with onboard data acquisition hardware. Data is typically transmitted from the LiDAR sensor to the computer through USB 3.0 or Ethernet cable which may vary with the manufacturer. For instance, the Velodyne-Alpha Prime model, providing ultra-high resolution 3D point clouds of the surrounding environment, transmits data using Ethernet connection: 1000BASE-T (IEEE 802.3ab) with a bit rate of 125 megabits per seconds [27]. Ethernet user datagram protocol (UDP) and transmission control protocol (TCP) are the basic communication protocols that are generally used.
The computer on receiving the data acquired by the LiDAR sensor, must be able to visualize and process it. Manufacturers provide graphical user interface (GUI) to visualize and convert the the real-time raw sensor data into a machine-readable format along with software development kit (SDK) to customize the sensor data acquisition. These GUIs and SDKs are unique to the manufacturers of the LiDAR and vary from one manufacturer to the other. An example of one such GUI is the VeloView by Velodyne. It performs real-time data visualization of live streamed 3D LiDAR data from Velodyne's sensors, namely, HDL-64E, HDL-32E, VLP-32, VLP-16, Puck, Puck LITE, Puck Hi-Res, Ultra Alpha, Alpha Prime, Velarray, Velabit, Veladome. VeloView also provides raw sensor data in.CSV format and custom color maps of intensityof-return, time, distance, azimuth, dual return type, and laser id for post-processing. Manufacturers also provide post-processing workflow platforms. These platforms provide additional functionality of post-processing of the collected raw data as compared to GUIs. A popular example of such workflow platform is HxMap by Leica Geosystems. The workflow of HxMap involves data pre-processing followed by geometrical and radiometric data adjustment followed by product generation (LiDAR 3D models or customized templates). Leica Cyclone Cloud is another cloud-based software service by Leica Geosystems [28].

C. DIFFERENTIAL ABSORPTION LiDAR (DIAL)
Differential absorption LiDAR (DIAL) is a laser remote sensing method that is applied for measurements of atmospheric gas concentrations. DIAL employs two different frequencies to measure the scattered light through a medium. Each wavelength interacts differently with the medium; one of the wavelengths gets attenuated more than the other. Thus, the back-scattered light from each of these wavelengths is different and can be used to compute the concentration of different constituents of the medium. The DIAL has proven to be useful to measure tropospheric concentrations of NO, H 2 O, O 3 , SO 2 , and CH 4 , as well as ozone levels [29]. DIAL is highly relevant for infrastructure monitoring as it can detect hydrocarbon concentrations [30], allowing the inspection of pipelines covering big distances in short amounts of time.

D. FLUORESCENT LASER SPECTROMETER (FLS) LiDAR
Fluorescent laser spectrometer (FLS)-LiDAR uses hyperspectral detectors to collect laser data. This allows subsequent data analysis systems to determine qualitative and quantitative analyses of both water bodies and land surfaces, offering an integrated environmental assessment based on airborne surveys [31].

E. LiDAR DATA SETS
The number of available LiDAR data sets over the last few years has been steadily increasing. These data sets have files with large amounts of data. Usually, airborne-based LiDAR data sets are subdivided into map regions to break down the data into smaller geographical areas. The Canadian LiDAR data [32] presents LiDAR data sets from Canada, open for free download and download with restrictions. Most of the data sets, if not all of the data sets listed in [32], have raw LiDAR data collected from airborne systems, but they do not have data annotations for infrastructure classification tasks using supervised ML approaches.
The OpenTopography website [33] provides a list of data sets from around the world. The user can go through a list of websites or zoom into a world map that is available on the website and select the region of interest. Normally, the OpenTopography website includes data sets from a variety of streams: airborne LiDAR, terrestrial laser scanner (TLS), bathymetric data, satellite data, and structure from motion/photogrammetry. Some of the data sets even allow comparison between measurements collected on different occasions. Some of these data sets were collected after natural disasters, such as the hurricane Harvey in 2017, to evaluate the magnitude of the flood, determine the probability of occurrence, and map the extent of the flood in Texas [34]. Data recorded from environments after disasters can be valuable for monitoring the safety of infrastructures that are crucial for relief operations and safety of the people in the affected regions. For example, may be identified and monitored for maintenance and safety purposes. Further identification of specific areas where maintenance is required can also be very valuable under such circumstances.
Airborne-based LiDAR data sets for semantic segmentation can also be found in the literature. The DublinCity data set presented in [35], available for download in [36], covers around 2km 2 of annotated point clouds. Fig. 5 (red boxes around the annotated areas) and Fig. 6 (the class hierarchy) show an overview of the DublinCity data set. The LiDAR data was collected from a helicopter airborne laser scanner (ALS) system and includes around 260 million labeled points.  [35], [37].
The Dayton annotated laser earth scan (DALES) data set, presented in [38] and available for download in [39], consists of point cloud data collected from ALS, with around 5 billion hand-label points over 10km 2 and has 9 major categories: VOLUME 11, 2023 FIGURE 6. Class hierarchy [35], [37]. ground, vegetation, cars, trucks, poles, power lines, fences, buildings, and unknown. Fig. 7 shows color-coded tiles for the different classes available on this data set. The intersection over union (IoU) is a common metric to evaluate a machine learning model on point cloud classification. The authors in [38] reported an IOU of 95.5% for power lines and 75% for poles. Annotated power lines, poles, and buildings are useful for monitoring and classification of critical infrastructure areas such as energy, utilities, and safety.
Initiatives for semantic segmentation using terrestrial data can be found in the literature [40], [41]. The Toronto-3D data set presented in [40] is available for download at [42], while the Paris-Lille data set presented in [41] is available for download at [43]. The Toronto-3D data set includes the 3D data points collected using a vehicle-mounted system, and has 8 labeled classes: road, road marking, natural, building, utility line, pole, car, and fence. This type of data set is an important step towards semantic segmentation from LiDAR data, which could enable automated monitoring of classes of interest, including critical infrastructure.
Semantic segmentation applied to self-driving vehicles can also be found in the literature. Table 2 shows a few data sets available in the literature that uses LiDAR for object segmentation (listed in [44]). While these data sets are focused on important objects for self driving cars, classes such as pole ( [45], [46]), traffic sign ( [45], [46]), traffic light ( [46]), bridge ( [46]), tunnel ( [46]), and tree ( [47]), the information extracted from these data sets can be used for infrastructure monitoring and can provide useful information for activities such as planning and maintenance.
In summary, LiDAR is an emerging sensing technology that rapidly and inexpensively builds a 3D model of the scene. It is safe to use near humans and is versatile in being able to survey multiple surfaces (e.g., both land and water). LiDAR can be mounted on a moving vehicle (e.g., plane, train) or in a fixed location. These qualities make LiDAR suitable for continuous monitoring of a wide range of critical infrastructure, as detailed in the following section. The availability of rich data sets enable the study of critical infrastructure protection viable and possible.

III. APPLICATION AREAS
In this section, different applications of LiDAR, such as in monitoring road, rail, and air transportation, are studied. LiDAR-based approaches in monitoring critical infrastructures of pipelines and electrical power transmissions and distribution, mining, and nuclear plants are also reviewed. Besides, detection and tracking of people and safety of critical areas for security purposes are also discussed.
The following performance metrics are used in the evaluation of various methods in the literature: • Accuracy (correctness): the proportion of correct predictions (classifications) among the total number of cases examined.
where TP is true positive, TN is true negative, FP is false positive and FN is false negative.

A. ROAD TRANSPORTATION
Transportation unifies a nation, links various cultures of a nation, and strongly influences a nation's economy. Any interruption in the cycle of transfer of goods and services leads to considerable losses to a nation's economy. The recent emerging technology, such as self-driving cars, acquire and use information about the infrastructure. For instance, autonomous driving systems are required to have the ability to recognize the various objects on and to surround the roads. We review some of the modern LiDAR-based methods to monitor transportation infrastructures. LiDAR-based technology helps to extract information about road surface geometry and other existing objects in the roadway. The following section elaborates the application of LiDAR technology in road infrastructure monitoring.

1) ROAD EDGE EXTRACTION
Road edge extraction provides us with many data for different utilization, specifically recognizing the road's potential danger factors. For instance, if some parts of the road environment are defective, it can mislead a vehicle driver or cause a severe crash. In [50], in order to extract street characteristics from 3D point clouds captured by terrestrial mobile LiDAR, a segmentation approach was proposed using an active contour model. The road surface information was extracted using the edge detection method on the LiDAR data set. Methods for extracting road surface focus on extracting the curb edge. In the active contour method, the road and the curb point information are required to be different heights. To this end, the intensity and pulse width properties of LiDAR, along with the data set were used for extracting these information. Fig. 8 shows the extracted 3D edges using LiDAR data.

2) ROAD GEOMETRIC PROPERTIES EXTRACTION
It is necessary for transportation control safety purposes to have access to reliable data that contains the road's geometric properties such as a slope gradient, travel width, and curve radius. According to [51], geometric properties are detectable by applying a 3D auto-detection technique using LiDAR data.
Usually, 3D auto-detection methods use different devices like Leica Total Station, a known industrial system brand in measurement technology that uses 3D laser scanning techniques to provide input data. With emerging LiDAR technology, the use of LiDAR as a supplier of input information for 3D detectors is entirely reasonable. Because, by using the LiDAR point cloud information, it is possible to detect the vertical distribution of the object on the ground. More specifically, for the forest road, LiDAR is able to detect ground points below the masked surfaces. Hence, nowadays, LiDAR systems are counted as strong tools in forest road monitoring [51].
A study in the north of Iran showed that by using airborne LiDAR, there is less average error in forest road vertical and horizontal profile assessments than assessments achieved by Leica Total Station data. The results of this study showed that geometric properties of forest roads can be obtained even under dense plants canopy by using high-resolution LiDAR data [51].

3) ROAD SURFACE FEATURES EXTRACTION
Road surface features, like road marking, play an important role in driving safety. Such features can be utilized to develop autonomous vehicle assistance. In order to facilitate such a development, it is necessary to develop a trusty database of road surfaces [52]. Such a database should contain road surfaces under different climatic conditions,parameters that influence the condition of the road such as number of vehicles that are passing on the roads, vehicle size and weight, the condition of the road as delivered by the status of maintenance of the road (for instance types of repairs).
It is possible to reconstruct a road surface automatically using the road surface features obtained from the data collected by LiDAR. The impact of ambient light levels on LiDAR is low, and LiDAR can provide easily provide highquality data. Fig. 9 shows the results of an experiment conducted on a LiDAR-surveyed route in Taiwan and illustrates its LiDAR point cloud data. Results obtained using terrestrial mobile LiDAR in the experiment demonstrated that the method suggested in [52] for road surface reconstruction was able to obtain an overall classification accuracy of 90%.

4) ROAD SURFACE, CENTERLINE AND BOUNDARIES DETECTION
Road surface and its centerline and boundaries information are used for several purposes like roadway network development and engineering. In [53], it has been shown that centerline and boundaries can be detected accurately by first converting the LiDAR data to an intensity image and then by identifying the candidate pixels. By using terrestrial fixed LiDAR, more than 90% completeness and correctness have been achieved. Fig. 10 illustrates reconstructed LiDAR points of a road surface.

5) ROAD DETECTION IN AUTONOMOUS DRIVING SYSTEMS
Road recognition is a crucial procedure for autonomous driving systems. Without reliable detection of road sections, a self-driving vehicle cannot be safe. A method to improve detection based on LiDAR data was developed in [54]. Commonly, the road detection method confronts problems that arise because of light variation and blurry images. LiDAR data is less sensitive to visual turbulence. Hence, LiDAR utilization is growing significantly in this area. Using the terrestrial mobile LiDAR data, a method called Progressive LiDAR adaptation-aided road detection (PLARD) has been developed on the KITTI benchmark. This method shows considerable robustness and promise as a better road detection method than other state-of-the-art road detection techniques.
Normally, many autonomous driving vehicles use LiDAR devices to recognize road lanes. However, a sizable number of them are just able to detect uncomplicated roadway types. A method to detect all road features and parameters for complex streets with many lanes and signs like arrows, stop signs in real-time was proposed in [55]. In this method, input data was provided by a terrestrial mobile LiDAR device. Results demonstrated the reliability of the 3D LiDAR point cloud as an input database for real-time complex roadways detection. Fig. 11 shows three examples of road detection in complex routes by this technique.

6) ROAD DETECTION IN ROAD TRAFFIC MONITORING SYSTEMS
Today, road traffic monitoring system (RTMS) is one of the most practical systems that modern cities are using. RTMS is designed to improve vehicle traffic flow and enhance street safety. Traditional algorithms that use cameras for monitoring suffer from environmental 'noises' like shadows and bent lines. LiDAR sensors are not affected by ambient noise parameters. A decision-tree roadway recognition approach introduced in [56] used a fixed terrestrial LiDAR sensor for data acquisition. The LiDAR was able to scan all the 3D targets in the intended environment and provide point cloud data. This method was able to provide broader roadways recognition with less detection time.

7) ROAD CRACK EXTRACTION
Several parameters like extreme loading, fatigue, rain, and thermal fluctuations can cause roadway cracks. Ideally, it is essential to monitor asphalt surfaces continuously to find new cracks as soon as possible. Cracks on the road may lead to vehicular damage and accidents. Pavement cracks normally show lower intensity of point clouds than their surroundings. The L1-medial skeleton method extracts cracks based on the clustering method. Results validate the effectiveness of the introduced algorithm using terrestrial mobile LiDAR point clouds [57].

8) ASPHALT PAVEMENT DISTRESS IDENTIFICATION
It has been emphasized in [58] that the LiDAR database effectively helps in monitoring road health conditions. In the case study in China that was presented in [58], a random forest classifier technique was applied to airborne LiDAR data and more than 95% accuracy was achieved in recognizing road distress. In comparison to digital images, LiDAR can directly obtain the 3D spatial data of degenerated asphalt to measure the geometric dimensions of the pavement issues.

9) STREET TRAFFIC SIGNS AND LIGHT POLES DETECTION AND CLASSIFICATION
Traffic signs play a critical role in cities' transportation networks because they help to regulate traffic, guide drivers and provide warning to the drivers. They convey advance warning to the drivers and thus help them in responding to the conditions on the road. Thus, controlling and managing their condition and serviceability are paramount for maintaining safer roads. In [59] a traffic sign classifier was presented using a LiDAR data set. Typically, traffic sign detectors are designed mostly based on digital images and video data. Image-based or video-based information is affected by many conditions such as shade and weather. Weather conditions like fog or rain and shadows created by objects in the neighbourhood of the traffic signs influence the traffic signs' visibility. On the other hand, laser scanning systems, like LiDAR-based technologies, are generally not affected by the above mentioned conditions. LiDAR-based technologies also provide very dense and precise (3D) point clouds. By using terrestrial-mobile LiDAR and digital data, the classifier used in [59] achieved an accuracy of 86.8%.
Using the data acquired with terrestrial mobile LiDAR, in [60], clustering technique was used to recognize street light poles. An accuracy of 97% was reported. Fig. 12 shows results for light pole extraction in the street.

10) DETECTION OF NON-PHOTOSYNTHETIC COMPONENTS OF ROADSIDE TREES
Roadside vegetation are significant elements of environment and ecosystem. More specifically, vegetation reduce pollution and contributes to ambient beauty. Non-photosynthetic components affect the growth of vegetation and may also contribute to fires. Thus, detection of non-photosynthetic components is crucial for safety of roads and road traffic.
In [61], a cluster-based approach was proposed to identify non-photosynthetic parts of roadside trees. This method uses terrestrial mobile LiDAR point clouds and robustly recognizes non-photosynthetic components with a hierarchical clustering strategy. Table 3 briefly compares the various roadways applications using LiDAR-based technology. As shown in the table, terrestrial LiDAR have been used extensively in almost all the research work in the literature.

B. RAIL TRANSPORTATION
LiDAR-based technology provides spatial information regarding rail geometry and other objects in a railroad network. Typically, high-resolution data is required for monitoring railroad networks. To this end, research findings show that LiDAR collected data represents an appropriate and reliable source for a number of rail infrastructure monitoring tasks.
In this section, a brief review of applications using LiDAR in the monitoring of railroad network infrastructures is presented.

1) RAILROAD CENTERLINE EXTRACTION
Railroad centerline extraction is used for checking the required clearance near rail tracks. Therefore, it is necessary to have exact data about rail track geometry. A model for automatically extracting railroad centerline using LiDAR point cloud (input data with high accuracy) was proposed in [62]. In order to extract railroad centerline features, two approaches for the modeling part were presented. In the first approach, center locations were created by projecting points of the railroad. This was done in a data-driven way by duplicating railroad track points to the parallel track. Then the midpoints were considered as the primary centerline and  a piece-wise linear function is used to produce center points at regular intervals. In the second approach, rail track was formed by implementing piece-wise 3D track patterns to the rail track points. Fig. 13 illustrates a 3D model of the railroad and its centerline and point cloud data. VOLUME 11, 2023 2) RAILROAD ELEMENTS EXTRACTION Preventive actions like gauge monitoring, ballast assessment, and surrounding slope maintenance are vital in railway infrastructure monitoring. To this end, accurate information regarding a railroad and its surrounding objects' geometry is needed for ensuring train circulation safety.
According to [63], the 3D linear elements in railroad infrastructure are detectable using LiDAR point cloud features. Usually, gathering data can be done by a LiDAR device directly mounted on the train or installed on a car carried by train. In [63], the LiDAR platform was positioned directly on the train and a semi-automatic method for 3D linear element extraction was carried out. Metrics such as completeness, correctness and quality were defined and used to evaluate the performance of the proposed method for centerline geometry property extraction.

3) RAILROAD INFRASTRUCTURE RECOGNITION
Railroad infrastructure comprises rail tracks, contact cables, catenary cables, return current cables, masts, and cantilevers. In order to address the safety concerns in the railroad transportation system, continuous monitoring is recommended.
Traditionally, it is common to use a visual inspector to identify problematic parts in railroad networks. 3D LiDAR data technology is a better choice as it is cheaper and its performance is superior to a visual inspector.
A classification algorithm implementation on an Austrian rural railroad [64] showed that it is possible to recognize a railroad system's main components based on their physical form, geometrical characteristics, and topological relationships. Fig. 14 depicts a sample LiDAR installed on a train, and Fig. 15 illustrates the recognized railroad infrastructure. Classification metrics obtained with a terrestrial mobile LiDAR demonstrated that automatic recognition of the railroad infrastructure method is robust [64].

4) RAILROAD INFRASTRUCTURE DETECTION
Some notable features of LiDAR technology have led to its increased usage of infrastructure monitoring in recent years. The beneficial attributes such as almost no human intervention (therefore minimum workforce), ability to gather a large amount of data and cover a large area, ability to provide non-stop continuous detection, and to provide the ability to analyze historic data for any investigation have led to LiDAR becoming a preferred sensor for railroad infrastructure detection. For instance, it was demonstrated in [65] that by using railroad path geometry and iteratively applying a vertical slice to point cloud data, it is possible to detect railroad network structures in-line and evaluate their dimensions and spatial location. By using terrestrial mobile LiDAR, the method demonstrated in [65] achieved 100% accuracy in railroad structure detection.

5) ON-BOARD RAILROAD DETECTION AND TRAIN LOCALIZATION
Normally, train localization methods present a sensor configuration that connects the global navigation satellite system (GNSS) with a velocity sensor and a digital track map. The major issue is that GNSS may not be available in sections like tunnels and under bridges [66]. With LiDAR as the sensor, localization can be achieved without the use of GNSS and lighting.
A template matching approach on LiDAR data was presented in [66] to improve object detection as LiDAR sensors are not affected by external lighting. Detection and localization can be achieved even in tunnels and under bridges where lighting is an issue. LiDAR sensors are large enough to distinguish main path, nearby roads, and objects next to the rails. By integrating terrestrial mobile LiDAR data into the navigation system, in [66] LiDAR sensors were used to detect railroad components and help the operators find the exact position during a loss of GNSS signal situations. Experimental results with a prototype system for a light rail scenario showed that the proposed on-train LiDAR-based localization system was notably robust and tolerant.

6) RAILROAD POWER LINE EXTRACTION
The railroad power supply is a critical subject in safety evaluation. Power loss may cause a railway transportation disaster.
Thus, monitoring and management of railway power lines is a necessary activity of railway maintenance.
Using the data acquired from RIEGL VMX450 terrestrial mobile LiDAR system, an automatic method to extract power lines was proposed in [67]. In this work, piece-wise segmentation of LiDAR point cloud data based on self-adaptive region growing method was used to extract power lines parallel with rails. Further, junctions were identified from the extracted power lines based on thresholds from information entropy evaluation.
An analysis result containing railroad point clouds to detect power lines is presented in Fig. 16.

7) RAILROAD SURFACE QUALITY MONITORING
Usually, expert maintenance staff evaluates railroad quality concerns visually. An analysis of the top of rail lubricity status and layer body using the Terrestrial LiDAR data set was given in [14]. A shift in Doppler frequency spectrum was observed due to rail structural irregularities and dissimilarity due to the presence of foreign objects on the rails. This demonstrated the potentials of LiDAR sensor to discover and quantify top of rail roughness, friction modifiers, and crushed leaves on the rails. This study confirmed the applicability of LiDAR sensors for monitoring of railroad network infrastructure.
LiDAR also serve as an excellent auxiliary sensor for enhanced railroad structural health monitoring. In [68], a multi-sensor system consisting of LiDAR and multi-temporal interferometric synthetic aperture radar (MT-InSAR) has been introduced to monitor railroads' structural health. Taking advantage of both radar scatterers and LiDAR point clouds can resolve the ambiguity that exists while using only a radar scatterer.

8) RAILROAD CLEARANCE CHECKING
Obstructions on railroads represents significant dangers in rail transportation systems. Consequently, it is crucial to recognize objects, like encroaching trees by a clearance gauge procedure. Currently, human operators monitor the upcoming track clearances from the engine at the front of the train; however, this is a cumbersome task over prolonged journeys. Therefore, railroad clearance checking would benefit  from automation using forward-scanning LiDAR and suitable processing.
According to [69], by using iterative closest point algorithm, the clearance in rail track was estimated automatically and a rail clearance of about 3mm was reported. Fig. 17 displays an example of the obtained point cloud data by LiDAR for use in clearance gauge. Table 4 compares the various railways applications and research work carried out using LiDAR technology that are currently available in the literature. VOLUME 11, 2023

C. BRIDGES 1) RAILROAD BRIDGE DAMAGE ASSESSMENT
Many factors like excessive loading and aging can cause railroad bridge damage. LiDAR technology is able to provide maintenance authorities with useful information for monitoring purposes.
In a case study by [70], a bridge in the UK was investigated in order to study its historic deformation. Point sensors, which are fixed inside of optic fiber cable, provide monitoring capability against train weights. Results show that LiDAR point cloud data can foretell possible settlements to an accuracy of almost 0.5 cm [70].

2) HIGHWAYS BRIDGES DAMAGE EVALUATION
High-quality terrestrial photos produced from laser scans supply precise geometric data regarding a structure, from which damaged areas can be identified and measured. Terrestrial mobile LiDAR has been shown to be useful for bridge damage condition evaluation.
In [71], LiDAR scans were used for quantifying damage on large infrastructures. By using LiDAR point clouds, it was demonstrated that small damages can be quantified and cracks can be captured. Several case studies such as mass loss quantification, damage quantification comparisons of bridges using distance and gradient combined threshold method and minimum mean square error and triangulation method, crack detection and evaluation were conducted. Using 3D LiDAR, evaluation can be conducted without stopping the traffic and existing conditions can be recorded and preserved for eternity.

D. INTELLIGENT TRANSPORTATION SYSTEMS (ITS)
Intelligent Transportation Systems (ITS) refers to a collection of advanced information and communication technologies that are utilized in traffic management and transportation systems to increase the reliability, sustainability, and safety of transportation networks, to ease traffic congestion, and to improve the driving experience. ITS using LiDAR can detect, locate and measure the distance between the ego vehicle and nearby vehicles, pedestrians and cyclists to generate data.
Robots with onboard LiDAR are operated remotely in the field of ITS. It may be beneficial to send 3-D point clouds in place of two-dimensional images as two-dimensional images are not capable of providing depth information. By using LiDAR, data sets described by point clouds consist of 3D points accompanied by additional information such as color and distance. As a result, the utilization of LiDAR in the ITS industry is on the rise.

1) REAL TIME LiDAR DATA COMPRESSION AND TRANSMISSION
LiDAR sensors are used in applications to find range and depth information of transportation systems by actively illuminating using a laser and acquiring the mirrored pulses. Such applications are becoming more widespread in the the domain of surveying and ITS. The size of the point cloud data produced by LiDAR can be significant.
In [72], a real-time point cloud transmission over Wi-Fi is presented. To reduce data transmission requirements, an Octree-based point cloud compression method is employed to avoid the transmission of massive data. This compression approach is implemented by the point cloud Library (PCL).

2) DETECTION OF GNSS NO-LINE OF SIGHT SIGNALS
Vehicle positioning on land, at sea, and in the air is most commonly carried out using GNSS technology. In the GNSS, a receiver's absolute position is calculated in relation to a fixed reference. GNSS systems are comprised of a number of satellites orbiting about 20,000 kilometers above the earth's surface. Satellites emit signals containing information about their location, orbital parameters, etc. Those signals are received by the system's reception systems, which are capable of extracting information regarding the location, speed, and time of the signals. GNSS provides positioning, navigation and timing (PNT) information to support the ITS globally or on a regional basis. However, GNSS cannot provide PMT information with same accuracy in all parts of the world due to no-line of sight (NLOS) signals, especially in the metropolitan areas.
In [73], the integration of GNSS and LiDAR for the detection of NLOS signals was presented. The results demonstrated an enhancement in the horizontal direction's positional accuracy after using the presented algorithm which was based on detecting NLOS from the search strip obtained using the LiDAR point clouds and the position information available from the GNSS. Real area information was employed to control and validate the suggested procedure and method.

E. AIR TRANSPORTATION
Apart from providing smart solutions for passenger luxury and streamlining the daily airport operations, smart solutions to ground surveillance and security issues also play an important role in developing smart airports. LiDAR can significantly contribute to providing such solutions in the following ways:

1) FOREIGN OBJECT DEBRIS (FOD) DETECTION
FOD can be defined as any item of any size or material found in an inappropriate location within an airport runway, taxiway, or ramp areas such as aircraft parts, nuts, bolts, luggage items, aircraft handlers' personal items, and construction debris. The presence of FOD can cause significant damage to aircrafts and airport equipment, as well as induce injury on operating personnel on the airfield. For example, FOD objects could be absorbed by the jet engines causing damage to turbine blades.
FOD relies heavily on human power and the labor force. Several humans are employed to keep the airports free of FOD. This practice is subject to human errors and possible mishaps. An autonomous rover that uses LiDAR technology to scan and detect foreign objects on airport runways or ramps is being used. A rover is cheap to implement and is designed so as not to interfere with airport airside operations. The main subsystem in the autonomous rover is a sensor that utilizes LiDAR technology for locating the foreign debris [74].

2) MONITORING AIRCRAFT MOTIONS
A number of accidents take place due to aircraft collisions on the ground, making safety in air and land equally important. Obtaining reliable and accurate navigation information from landing and taxiing aircraft is difficult. Airport authorities use radar to monitor airplane traffic, which provides only coarse positioning and almost no attitude information. Imaging sensors are especially effective to acquire a large volume of data to extract objects, such as aircraft bodies, and then track them based on image sequences [75]. In particular, LiDAR is a suitable remote sensing technology for this type of data acquisition, as it directly captures 3D data and is reasonably fast. If the object size/shape is known, this information can be beneficial to estimate motion parameters, such as velocity.
The use of a new LiDAR waveform deconvolution and georeferencing strategy that produces very dense, detailed point clouds in which the vertical structures of objects are well characterized is the latest way of processing the acquired data [76].

3) DIGITAL TERRAIN MAPPING
The safety of the aircraft and airline passengers relies heavily on the safe and efficient use of the air space in the vicinity of an airport and en route to the airport. The airport operating agency is responsible for ensuring that the aerial approaches and the airfield area are adequately cleared and protected. Terrain topography and urban growth are important factors, which influence the efforts required by the airport operating agency to make these decisions in the most cost-effective way.
LiDAR systems built upon the Optech ALTM laser technology are used in the LiDAR survey projects and digital terrain mapping. Flight-path planning is another important factor in the LiDAR system mission. The flight path shall cover the study area satisfactorily including both parallel and enough cross flight lines to eliminate shadowing and allow for proper quality control [77].

4) WEATHER ANALYSIS
There has been a significant increase in the air traffic over the last three decades, leading to a demand for better air safety measures. Adverse weather conditions and wind hazards pose a major threat during the landing and take-off of the aircraft. Abrupt changes in wind speed and wind direction, inadequate spacing of wake vortex separation, and gust measurement may dramatically influence the aerodynamic circumstances of the flight and may threaten the safety of the aircraft, causing about half of the air navigation system (ANS)-related accidents globally. Weather LiDAR prove to be a promising technology in addressing these threats in the following ways:

a: WAKE VORTEX MONITORING
Wake vortex is a disturbance in the atmosphere that forms behind an aircraft as it passes through the air. Accurate remote detection and tracking of aircraft wake vortices along approach and departure corridors have gained significant interest in the past several years. Adequate separation time between aircrafts can reduce the probability of aircraft catastrophe. In nearly all cases, the sensor is set up to view the aircraft wake vortices with a vertical scan perpendicular to the aircraft flight track. This is called the transverse viewing geometry.
Pulsed Doppler LiDAR mounted on-board an aircraft are being extensively used for remote sensing of wake vortices. The LiDAR that is extensively used for wake vortex monitoring by the Japan aerospace exploration agency (JAXA), radiates 1543 nm laser pulses with a physical length of almost 50 m. These emitted pulses are scattered by aerosols flowing with air, and back-scattered light is received and digitally sampled. For each direction, sample data of 600 pulses at a pulse repetition frequency of 18 kHz are processed and accumulated, and measurements of carrier-to-noise ratio (CNR), mean radial velocity, velocity dispersion, and Doppler spectrum are output for each 5m range bins to a distance of nearly 900 m. Fig. 18 illustrates a sketch map of airport wake vortex monitoring. FIGURE 18. Sketch map of airport wake vortex monitoring [78]. VOLUME 11, 2023 b: WIND SHEAR DETECTION A terminal Doppler weather radar (TDWR) has been used to detect and warn low-level wind shear. While the TDWR has proved to be effective in rainy weather, reports received from aircraft pilots landing at or taking off from the airport indicate that low-level wind shear also occurs under clear-air conditions. In order to overcome this shortcoming, a Doppler LiDAR system was installed in mid-2002 to monitor wind flow around the airport [79].

c: GUST AND MICROBURST MEASUREMENTS
Gust fronts are elongated regions of converging flow and are often associated with a thunderstorm outflow but may also result from downslope wind conditions, cold fronts, sea breezes, and so on. An aircraft flying across a convergent flow region in either direction will experience a lift due to the apparently increased headwind. Microbursts are typically produced within thunderstorms and are characterized by an area of strong diverging surface flow that results from a significant downdraft from a thunderstorm cell. An aircraft encountering a microburst or similar negative shear will first experience an increasing headwind, which will increase the lift. However, soon after encountering the increased headwind, the aircraft encounters an increased tailwind after passing through the center of the downdraft (core). This causes a loss of lift and, unless properly countered, a dangerous and potentially fatal loss of altitude [80].
Coherent pulse-Doppler LiDAR serves as a solution to the problem discussed in the previous paragraph. Installationdependent scans are conducted, and radial velocity data is collected along each averaged line of sight in the scan. The radial velocities quantify the wind speed component along the sensor line of sight. If the wind profile along a line of sight is sufficiently divergent by an amount consistent with adjacent lines of sight, the region is declared a wind shear or microburst.

d: THE LATEST TECHNOLOGY: A COHERENT ALL-FIBER, MODULAR, COMPACT WIND LiDAR
Currently, the most widely used commercial LiDAR system in the world is WindTracer LiDAR systems. Organizations including national aeronautics and space administration Langley research centre (NASA LaRC) have used the data collected from WindTracer systems for validation of fast time wake models [81]. However, commercially available WindTracer or custom-built LiDAR systems have several disadvantages such as they are (a) large and heavy, (b) subjected to high-maintenance, (c) suffering from reliability issues, and (d) often prohibitively costly.
In order to overcome the disadvantages of current systems, a long-range, versatile wind LiDAR with all-fiber architecture using commercial-off-the-shelf (COTS) components was developed for wind and wake vortex sensing applications. The LiDAR system, known as Windimager, was developed by Sibell Optics under the NASA small business innovation research (SBIR) funding with emphasis on enhancing range, sensitivity, system efficiency and reliability and simultaneously reducing size, weight, and power (SWaP) and maintenance time [82]. Furthermore, it incorporated a novel 3D hemispherical scanner that was integrated into the LiDAR system. Modularity, reliability due to the Telcordia-tested COTS components, lower maintenance costs, a predominance of line-replaceable units, and cost-effectiveness are the pros of this system. Besides ground-based operations, this wind LiDAR technology also provides a viable technical path for airborne platforms.

F. PIPELINES
There are an estimated 840,000 kilometers of transmission, gathering, and distribution pipelines in Canada [83]. These pipelines are distributed in populated and remote areas of Canada, with 450,000 km of local distribution lines, 250,000 km of gathering lines, 117,000 km of large-diameter transmission lines (4''-48''), and 25,000 km of feeder lines.
The use of pipelines to transport oil and gas is considered safe, reliable, and environment friendly [83]. However, in the long run, gas and oil pipelines will suffer from corrosion and weld defects [84]. The most common failures in gas pipelines are cracking and perforation, while for the oil pipelines, it is perforation [84]. Monitoring these pipelines is, therefore, important for the prevention of accidents and heavy environmental and personal losses [84].

1) PIPELINE LEAK DETECTION
Monitoring gas pipelines can use many methods, including active methods such as LiDAR, diode laser absorption, and millimeter-wave radar and passive methods such as thermal imaging, spectral imaging, sound wave, acoustic emission, ultrasonic guided wave, ultrasonic, and sonar, but a combination of different methods can increase the accuracy of the leak detection [84]. Active methods can be used to transmit signals above the pipeline and then measure the amount of absorption for a given constituent [85]. If there is significant absorption, a leak on the pipeline can be assumed. DIAL can be used to determine the position of the leak within a 500 m range [86].

2) STRESS CORROSION CRACKING DETECTION
Monitoring the structural integrity of pipelines is important as a preventive measure. stress corrosion cracking (SCC) is a common problem where small cracks appear outside of a pipeline [87]. These cracks can lead to rupture and leaks over time. SCC phenomenon can be affected by multiple factors such as stress level, the concentration of corrosive species, and surface finish [87]. According to [87], trapped moisture between the outer wall of the pipe and the protective coating is the main cause of SCC. LiDAR can be used to determine soil drainage based on high fidelity terrain data, and therefore, identify stress on the pipelines that can possibly cause SCC.

3) AIRBORNE HYDROCARBON DETECTION
The authors in [30] presented a system to detect methane, propane, and light crude gas vapors during real-world data collections. The airborne natural gas emission LiDAR (ANGEL) project uses three different sensors installed on an airborne platform: a DIAL, a high-resolution mapping camera, and a color video camera. In their experiments, they flew 1000 ft above the pipeline at speed up to 150 mph, and collected data during simulated leaks. Figures 19, 20, 21 show examples of propane plumes and imagery collected during controlled tests simulating different leak volumes measured in standard cubic feet per minute (scfm) over different terrains. The ANGEL sensor was able not only to detect methane, the main chemical compound of natural gas, but also other hydrocarbons including propane, gas condensates, crude oil, and refined hydrocarbons like gasoline, aviation gas, diesel fuel, jet A, and kerosene.

4) WATER PIPELINE ASSESSMENT
While natural gas and oil leakage have more potential to damage the environment, water pipelines can cause significant damages to critical infrastructures due to erosion of the soil. LiDAR technology is invaluable for disaster assessment. The authors in [88] collected both pre-trench and post-trench data to evaluate airborne and terrestrial LiDAR measurements, and the reported bias and standard deviations indicate that it is possible to reliably detect ground displacements, which in turn can be used for detection of damage in some pipelines. In the event of an earthquake or large amounts of rainfall, airborne and terrestrial LiDAR can be used to compare the terrain during normal conditions and after the disaster to detect displacement hazards [88].

G. ELECTRICAL POWER TRANSMISSION AND DISTRIBUTION EQUIPMENT
Applications of LiDAR technology in power transmission and distribution system involves risky objects' examination along the power line, adequate automatic clearances between conductors and objects, fine measurement of the interval between conductors, transmission line 3D visualization management, topography change detection along corridors, re-rating, vegetation evaluation and management along transmission line [89]. LiDAR helps in route optimization, VOLUME 11, 2023  environment protection, reduced labor work, increased process quality and more profit to the clients. They are being extensively utilized in power line construction.

1) POWER TRANSMISSION LINE CORRIDOR CLEARANCE MANAGEMENT
The power line networks transmitting the electricity are located in numerous transmission line corridors (i.e., rightof-ways (RoW)) areas). The inspection of the RoWs is one of the major responsibilities of operating and maintaining power transmission grids. RoWs are vulnerable to potential hazards (e.g., icing), resulting in short-or long-term power outages. The ROW clearance anomaly detection measures the distance between the power lines and the surrounding non-power-facility objects in the corridor such as trees and buildings to judge whether the clearance is within the safe range. Traditional methods used for monitoring the high voltage (HV) transmission lines mainly rely on aerial-and ground-based human on-site inspections with optical measuring devices (e.g., telescope and video/infrared camera) to manually detect the defects such as faulty components or encroaching vegetation. These methos were automated using the LiDAR technology [90].
LiDAR acquires the spatial geometry of the detected object in the form of discrete 3D point clouds. ALS has become one of the primary information sources for the ROW inspection. The standard manned aircraft (e.g., helicopter) borne ALS LiDAR systems are limited because of the inflexibility and high costs of the fight acquisition. Unmanned aerial vehicle (UAV)s provide an alternative to the traditional remote sensing platforms. Using UAV LiDAR data to detect the ROW clearance anomaly can effectively improve the flexibility and mobility of the power line inspection and save a large number of manpower, material resources and time cost in the manned aerial vehicle or manual inspection. It is particularly important for transmission lines inspection in difficult terrains, such as mountainous regions [91]. The methodology involved uses the following steps: 1) Extraction of the power line point clouds, 2) Fitting the power line point clouds segments iteratively with 3D catenary curve model, 3) Combining the constructed 3D power line catenary, points of non-power-facility objects and the digital elevation model (DEM). The Delair DT26 LiDAR fixed-wing drone is perfectly fitted for long-distance corridor mapping of power lines [92].

2) POWER TRANSMISSION TOWER PARTS CONDITION MONITORING
Tower parts condition monitoring is mainly to monitor the local state, such as self-detonation, damage and flashover trace of insulator string, various kinds of wire clips and anti-vibration hammer damage, defect, deformation movement and damage of lightning protection facilities. The UAV LiDAR is used to collect high-quality image information near the tower.
The latest development suggests the use of inspection based on cable inspection robot (CIR) LiDAR data. Robot inspection uses a cable inspection robot as a motion platform. According to different inspection objects, robot inspections can be divided into substation robot inspection and CIR inspection. Substation robot inspection is used to replace manual inspection to execute substation inspection work [93]. CIR is a power robot traveling along the ground wire, so it can reach rugged terrain, e.g., mountains, forests, and lakes, and it can carry out a fine inspection in these regions. Fig. 22 shows the mechanical structure of CIR. Besides, Fig. 23 and Fig. 24 presents a schematic diagram of data collection using CIR LiDAR system and the collected point cloud of actual lines.

3) POWER DISTRIBUTION EQUIPMENT SAFETY CONTROL
The electrical power distribution system is the part of the electrical network that distributes electric power for local use. It contains numerous poles which are built of wood. Most of these poles are placed in rural regions and are FIGURE 23. Schematic diagram of data collection using CIR LiDAR system [92].  surrounded by plants. Near the power lines, the presence of a tree's branch is a hazard and lead to power outage or interruption if it touches the overhead wires of the electrical distribution system. Therefore, it is necessary to have a plan for controlling the vegetation. To this end, LiDAR-equipped systems can play a useful role in this area. Using TLS and cameras, one can collect helpful images and point cloud data of the electrical distribution network. Then, by applying ML models to detect shapes in point clouds, one would be able to recognize electrical equipment and vegetation. This stage helps us to control and modify plants near the electrical poles.

H. MINING
Mining is of great importance to Canada as it contributes to the Canada's economy and industry. According to [94], Canada ranks in the top five producing countries for 13 major minerals and metals: first in potash; second in uranium and niobium; third in nickel, cobalt, aluminum and platinum group metals; and fifth in gold and diamonds. Monitoring mining sites and their surrounding landscapes are crucial for infrastructure resilience during all stages like exploration, mine-site design, closure, and reclamation.

1) MINE MAPPING
While topographic LiDAR have extra information from inertial measurement unit (IMU) and GPS data to improve the accuracy of data points, mapping of underground mines can make use of IMU, but often is not able to acquire reliable GPS signals. The authors in [95] presented a novel graph simultaneous localization and mapping (SLAM) optimization based on generalized iterative closest point (GICP) 3D point cloud registration using consecutive frames, consecutive key frames and loop frames. Experiments were conducted at an underground mine laboratory and were compared to the existing 3D laser SLAM method. With the use of LiDAR data, the 3D point cloud registration for mine mapping was optimized. Fig. 25 shows a comparison of scene mapping that can be applied to mine mapping, including the method implemented in [95]. Based on the assessment from the volume of the dumps, the authors in [96] concluded that the pits of Belchatow meet the requirements of the reclamation project. LiDAR was later used to analyze the vegetation growth of the reclamation site. This analysis enabled forest managers to evaluate vegetation growth in areas from reclamation sites, as well as check the risk of erosion and landslides [96].

I. NUCLEAR PLANTS
In Canada, 15% of the electricity is generated by nuclear power [97], [98], [99]. Nuclear power plants are critical infrastructures and their monitoring is essential for safety and security of a country. LiDAR systems embedded on UAV allows for monitoring nuclear power plants from a safer distance.

1) LiDAR AND MULTISPECTRAL IMAGERY FOR RADIOACTIVE WASTE SITE LOCALIZATION
Chernobyl is an example of a nuclear disaster. The accident that occurred in 1986 resulted in high levels of contamination. Radioactively contaminated materials were buried in radioactive waste temporary storage places (RWTSPs), also known as trenches and clamps, and these sites were poorly documented [100]. Today, many locations of RWT-SPs remain unknown. The authors of [100] collected data from LiDAR and a multispectral camera mounted on a UAV system over the Chernobyl area. The LiDAR data was used to generate a DTM, while a multispectral camera was used to classify the trees on the map. The combination of LiDAR point cloud and the classification of the trees were used for later classification of all the cloud points to create probability maps of radioactive waste sites. Fig. 27 shows an overview of the proposed method for detecting radioactive waste sites implemented in [100]. Briechle, Molitor, Krzystek, et al. [100] were able to localize the previously unknown 38 borings using the combination of LiDAR and multispectral imagery data. A classification accuracy of over 97% for DTM grid points was achieved. In addition, the F1 score for class trench was 0.76 for area #1 and 0.75 for area #2, but the authors in [100] expected a higher score as they speculated that the group errors were located in trenches that were not considered during manual labeling. Fig. 28 shows the classification of trenches and clamps, and conducted borings in the Chernobyl.

J. SECURITY
Although the detection and tracking of people as not considered as a direct application of infrastructure monitoring, it plays an important role in securing the infrastructure against unauthorized access. A significant amount of research has been done in this field using LiDAR. Thus, to highlight the potential application of LiDAR in security and broaden our scope of application, we have included this section.
LiDAR can be used for securing critical infrastructure against unauthorized access. Oftentimes, access to critical infrastructures is allowed only for a select group of people. Detection and identification of such personal are therefore important for keeping critical installations safe. LiDAR offers certain advantages when compared to commercial camera systems. Color cameras require good illumination sources and have a limited distance range. Different illumination conditions affect the quality of images of the objects captured by the camera. Fig. 29 shows an example of the same image of an individual obtained with different illumination sources. The figure shows the camera images having different shadings. Infrared cameras provide improved performance in darker areas when compared to color cameras, but do not work for longer distances. LiDAR allows data collection at longer distances than color and infrared cameras, provides information about the size of the objects and is less susceptible to dark environments. In the following, LiDAR application for detection and tracking people is presented.

1) PEOPLE DETECTION AND TRACKING
Person detection and tracking can be achieved using LiDAR systems [101], [102]. Fig. 30 shows the implemented system, the point cloud, intensity view, and range view of a scene, where a LiDAR was used to detect and track a moving person [101]. The system was able to be executed in realtime, and even when the person was running, there was no sight lost at any time.

2) ToF SENSOR -QVGA RESOLUTION
While most reported uses of LiDAR for security use a range finder with rotating parts, the authors in [101] and [102] used a flash LiDAR without moving parts, with the  resolution of 128 × 128 pixels. Texas Instruments (TI), a big provider and manufacturer of integrated components (IC), provides a LiDAR sensor (OPT8241) with QVGA resolution (320 × 240), with sampling rates as high as 150 frames per second, for as low as USD $26.69. While the price is only for the IC, it is clear that these sensors, and consequently systems, are becoming more affordable. TI recommends the use of the ToF sensor for the safety of critical areas and to count and identify people by extracting their skeletal structure [104]. These applications are highly relevant for monitoring critical infrastructure, and could be applied in sectors that are identified and prioritized by country.

IV. CONCLUSION
This survey has provided a survey of applications of LiDAR sensors in monitoring critical infrastructure. We have discussed different applications of LiDAR in monitoring road, railroad and air transportation systems, weather, bridges, ITS, pipelines, electrical power transmission and distribution equipment, mining, the safety of nuclear plants and also humans for security purposes. Transport agencies, governments, power stations, and construction companies all over the world have already started to embed such sensors into their streamlined processes to improve the infrastructure safety and to lower maintenance costs. Many countries have constructed efficient and robust railways with the help of LiDAR. The resiliency of railroad infrastructure is enhanced through improved risk prediction capability, enhanced infrastructure monitoring, and preemptive maintenance leading to avoid potential damages to the infrastructure. In order to realize an automated monitoring system, LiDAR sensors are mounted on trains to scan the railroads in a short amount VOLUME 11, 2023 of time. LiDAR based monitoring thus reduces human involvement in rail surveys and also minimizes human errors. LiDAR technology is also used for automatic ground surveillance and securities in airports. Weather LiDAR technology has been used to improve the safety of air traffic monitoring in adverse weather conditions. Passenger and aircraft safety can be enhanced through the creative use of LiDAR technology. LiDAR technology can provide maintenance authorities with valuable data for monitoring extreme loading and aging that can result in railroad bridge damage. Also, LiDAR sensors are used to discover range and depth data of transportation systems by active illumination by laser and obtaining the mirrored pulses. Different LiDAR technologies have been used in monitoring oil, gas and water pipelines for leak detection, stress corrosion cracking detection. The application of LiDAR in monitoring electrical power transmission and distribution equipment has been studied in detecting risky objects that may touch the power line. Further LiDAR can also help in identifying topography changes along corridors. Moreover, LiDAR has been used for monitoring mining sites and landscapes around it during exploration, mine-site design, closure, and reclamation. LiDAR systems embedded on UAVs make safe distance monitoring of nuclear plants possible. The use of LiDAR technology for infrastructure protection positively impacts the economy of the country and provides confidence to the public regarding safety.
The outcome of this research work shows that 3D LiDAR point cloud data provides a unique opportunity for automatic infrastructure monitoring applications. LiDAR systems are able to operate at high speed and provide point clouds that can be used to obtain accurate estimates and good classification results. Further, artificial intelligence and ML-based techniques have enabled advanced processing of LiDAR 3D point cloud data.
Integrators of LiDAR with other sensors are now readily available. As the cost of the LiDAR reduces, LiDAR may be used along with other stand-off sensors like radar to form an integrated multi-modal sensor. The sensed output of such an integrated sensor may be used with algorithm developed using artificial intelligence techniques to provide infrastructure protection. Such improved cost-effective sensors maybe available ubiquitously in the near future. Although LiDAR may not become the sensors of today but its day is soon to come.
Normally, LiDAR point clouds represent huge data sets, both in terms of total size and data acquisition rate; therefore, future work should focus on developing solutions for efficient processing of LiDAR data in batch and in real-time. Deep learning algorithms are showing great promise in semantic segmentation tasks from LiDAR and from fused RGB and LiDAR data, and should continue to be pursued.
Cost never stops research from advancing. On the one hand, efforts to reduce the cost of LiDAR are being taken, engineering is advancing the quality of sensors. From a technology perspective, LiDAR is moving toward quantumenhanced sensing. Quantum sensors such as quantum radars and quantum LiDAR are being researched all over the world and quantum illuminating sensors seem to be the direction in which sensing technology is forging ahead.
In conclusion, LiDAR will be the sensor of choice for infrastructure protection in the coming years. As LiDAR technology matures, multi-modal sensing strategy will dominate. When quantum sensors become readily available, quantum LiDAR and quantum radar will be part of the advanced sensing systems for infrastructure protection. He was the Director of the Ottawa-Carleton Institute for Biomedical Engineering (OCIBME), from 2020 to 2022. He is currently a Professor with the Department of Systems and Computer Engineering. He is the holder of two patents and two disclosures of invention. He is the author of more than 200 journal articles and conference papers. His research interests include signal and image processing, biomedical signal processing, pattern classification, and applied machine learning. He served as a Board Member for IEEE Canada, from 2010 to 2018. He was awarded the IEEE MGA Achievement Award, in 2012, and recognized for his IEEE contributions with Queen Elizabeth II Diamond Jubilee Medal, in 2012. IEEE Canada recognized his outstanding service through 2016 W. S. Read Outstanding Service Award. He is also the Chair of the IEEE Ottawa EMBS and AESS Chapters. He has been the North America Regional Director of the IEEE Consumer Technology Society, since 2021. He has been involved in organizing several successful IEEE conferences and has been a reviewer for several IEEE journals and conferences. F. A. KWAMENA (Member, IEEE) received the Ph.D. degree. Since 2012, he has been the Director of the Infrastructure Resilience Research Group (IRRG), an interdisciplinary group responsible for promoting analysis of all-hazards risks and threats to critical national infrastructure, collaboration between the faculty and industry, and professional training and research. In addition, he has been the Director of the Energy Infrastructure Security Division, Natural Resources Canada, since 2002, and worked with the Federal Canadian Government for over 30 years. He is currently an Adjunct Professor with the Department of Computer and Systems Engineering, Faculty of Engineering and Design, Carleton University. He is also a Special Advisor to the Dean of the Faculty of Engineering and Design, on the implementation of the Master's Program in Infrastructure Protection and International Security. He is also responsible for the establishment and ongoing operations of the Canadian Resources Infrastructure Resilience Nexus (CRIRN), a state-of-the-art research and development facility of a multi-disciplinary team subject matter experts, including engineers, cybersecurity, intelligence officials and policy analysts that proactively address all-hazard threats to critical energy infrastructure systems. In this role, he is responsible for directing and undertaking measures, and developing policies and initiatives focused on enhancing the security and resilience of Canada's Energy and Utilities Sector; one of the ten key economic sectors designated by the Government of Canada as critical to the economy and well-being of Canadians.