A Survey on Continuous Object Tracking and Boundary Detection Schemes in IoT Assisted Wireless Sensor Networks

With the new age of data innovation, the Internet of Things (IoT) proliferation has drawn enormous thought and has applied to help applications in different fields i.e., natural assurance, military observing, and industrial applications. WSNs are the essential segment of IoT for monitoring as well as tracking. The most preeminent applications provide confinement and identification of continuous objects i.e. wildfire, toxic gas, bio synthetics concoctions, and so forth. In the case of continuous objects such as fire and toxic gases are detected to identify the boundary of damage and alert teams for rescue efforts. It is also helpful in identifying safe paths for rescue. We have investigated various existing surveys that carried out different concepts associated with continuous object tracking and find out the deficit of boundary detection of object. In order to replete the present cleft of analysis, we have inspected various current state-of-the-art works on boundary detection of a continuous object that has yet not been added to the current writing. This paper presents an extensive overview of different continuous object tracking schemes which involve energy efficiency, boundary detection, communication, data aggregation, and network structural design in literature with the aid of featuring taxonomy. We summarized, compared, and classified these schemes along with their analysis and performance. Moreover, for further evaluation mechanism, strengths and weaknesses of these schemes are presented. Finally, various state-of-the-art open research challenges are identified. Moreover, there is a need to overcome these challenges through novel and reliable arrangements by the researchers.


I. INTRODUCTION
WSNs are basically consisting of large number of low powered sensing devices which are capable for sensing multiple events with limited amount of resources [1]. Remotely areas which are not investigated up till now due to its hazardous nature and inaccessible places WSN is found to be most effectual solution [2]. A novel paradigm Internetof-Things (IoT) [3] devices are emerging to build up the cost effective wireless sensor nodes that are connected with The associate editor coordinating the review of this manuscript and approving it for publication was Razi Iqbal . internet in sensing and monitoring processes. With the rapid and progressive technological development, IoT assisted WSN are widely applicable in environmental protection, industrial applications, military, habitat, agriculture sector, forest fire detection, health monitoring, seismic disturbance, volcanoes, earthquakes after shots stress detection, smart buildings, and predictive maintenance, etc. The WSN is the basic component of the Internet of Things. The combination of IoT and WSN tends to edge technology [4].
The smart sensing devices are densely deployed in a target area for surveillance of an event of concern connected by wireless media [5]. These sensors have the ability to sense data processed and exchange the data with neighbor(s), they also responsible for report sensed data back to the sink or base station (BS). Sink Node (SN) has capability to widely communicate with outside world i.e. Laptop. Recently, sensor node deployment plays considerable roles in military surveillance, traffic control, environmental monitoring, gas leakage detection, oil spills, battlefield and intruder tracking [6]. For tracking and monitoring the physical or environmental conditions, (i.e. temperature, movement, sound, vibration and pressure) sensor nodes are distributed in the target area. Sensors have power scarcity issue, derived by limitations of battery size and capacity, which have a great influence on extending the network life in terms of sensor energy utilization [7]. Therefore, it is considered as a significant issue for prolonging the network life and for many applications of target tracking [8]- [10].
The continuous object scatters over a large area i.e., wild fire, agricultural infections [11], toxic gas leakage and oil spill [12]- [14] in industrial applications. Continuous objects change their shape and size dynamically. The tracking mechanism initially detects and estimates object position and then continuously monitors it [15]. However, sensor nodes are low powered with limited sensing capabilities and there is a need to enhance the energy efficiency, network life time [16] and object monitoring and tracking [17]. Efficient report generation is an essential requirement for tracking the location of object in real time. In individual object tracking, the main concern is about how to predict and track the next location of the target. Secondly, what procedure should be adopted to notify the immediate tasking nodes in a huge area [18].
Object tracking and boundary detection is a challenging problem due to its speedy movement over time, increase in size, change in shape and split into multiple smaller continuous objects [19], [20]. Although, boundary detection is an efficient way to send data to sink as compared to normal detection phenomena in which number of sensing nodes send data to the sink and enhances the data traffic and communication cost. However, reliability of data is an essential requirement while estimating the boundary but failure of boundary nodes reduces the reliability and accuracy of boundary detection [21], [22].
Existing surveys explore the target tracking, network architecture strategy, node deployment, energy efficiency, gas diffusion models and different tracking techniques for continuous object. These surveys lack in presenting the object detection and boundary detection schemes. It is most significant concept for continuous objects detection that has not surveyed yet. To fill this gap, we investigate various current state-of-the-art schemes which consider both scenarios of object detection and boundary detection. The main contributions of this work are explored as follows; 1) We conducted a comprehensive study on existing surveys conducted on the topic. Table 1 presents a summary to highlight the need for this survey. 2) We explored different continuous object and boundary detection schemes and presented review of these schemes in literature. A taxonomy is presented to classify the recent studies in literature. 3) Next, we conduct an analytical review of different studies that are discussed in literature. We also discuss the comparative review of these schemes. 4) Finally, we categorize open research issues and challenges that should be resolved through novel and dependable solutions by the researchers. The rest of the paper is organized as following Section II, we have included taxonomy of continuous object tracking. Section III, provides analysis of the schemes and comparative discussion to highlight strengths and weakness of associated schemes. In section IV, we discussed recent developments and open research issues and challenges and at the last in Section V, we conclude our work.

II. LITERATURE REVIEW
This section presents taxonomy for different object tracking schemes as shown in Figure 2. These schemes further categorize into individual and continuous objects but we are focusing on continuous objects. We also categorize these schemes under three sections like object tracking schemes, boundary detection schemes and both object tracking and boundary detection schemes.
A. CONTINUOUS OBJECT TRACKING SCHEMES C. Zhong and M. Worboys provided a dynamic convoy tree-based collaboration (DCTC) approach for mobile target detection and tracking. This scheme constructed a tree base structure (convoy tree). It contains the sensor nodes that are in the region of mobile target. As the target object moves convoy tree is constructed dynamically for adding and removing sensor nodes in respected region. When a phenomenon region is detected by the nearby sensors nodes, these sensors cooperate with others sensors nodes for root node selection and constructed a convoy tree. Root node gathers data about boundary detection from others sensors and based on that information it used some algorithm to acquire the accurate information of target. Communication overhead occurs when the sensors collaborate with each other for root selection and sharing information. Most of the time this technique focuses on individual object tracking such as vehicles [27], animals and humans [28]. L. Liu et al., introduces a continuous object boundary detection protocol to effectively detect the faulty nodes with a minimum number of boundary tracking nodes [29]. J. H. Kim et al. presented the energy efficient Detection and Monitoring for Continuous Objects (DEMOCO) approach, it selected small number of boundary nodes, to minimize the message size transferred from boundary nodes to the SN, this algorithm use RN among BN for data transmission. If a node has the different reading from the previous reading, then that node become ''changed value node (CVN)'' and broadcasts COZ message to its onehop neighbors. When BN received a COZ message, it gets shorter back-off time and wake up early. It banned the other BNs to become RNs through control message. To lessen the message size, RN sends only the nearest neighbor's node ID among various neighbors of sink which carry unique values. For boundary identification this approach exchanges massive messages. DEMOCO approach is modified to examine the sensing range for Boundary Accuracy of continuous object tracking [30].
T. R. Sheltami et al. presented a Continuous Objects Detection and Tracking (CODAT) Algorithm for detecting and tracking the expanded and shrinking phenomenon. It also monitors holes inside the phenomenon. CODAT algorithm is hybrid of COBOM and DEMOCO algorithms. Through these algorithms energy consumption is gained by the help of few nodes selection among large number boundary nodes for reporting, but the boundary accuracy is compromised. CODAT removes the deficiency of boundary accuracy and used the average report data size by RNs for forwarding to sink. In this approach when a sensor node detected the phenomenon it broadcasts PCM after checking the current PDS status that not matched with previous reading. When other sensor node receives PCM, it matches its own PDS with it. If it comes to know that PDS is same then it will discard the PCM. But in the case of at least one PDS is different, and then this node becomes a BN. After that it counts the number of PCMs different waiting time and used this information for making RN. RN reports the boundary information with recent tags to the SN. During the expansion of continuous object only the nodes in the outer region of the phenomenon become BN. Correspondingly, in shrinking of object BN will be those nodes that are inside the region. This reduction of BN tends to energy efficiency as compared to COBOM. This technique minimizes the communication cost and provides the boundary accuracy but in sparse network reducing the sensor node cause problem while detecting the object and sensing holes [31].
Chengyue et al. introduces a novel architecture based on both static and mobile sinks. Mobility of mobile SNs based on the instructions of geographic information system. A centroid algorithm presents to optimally calculate the location of the mobile node through static node and both sinks collaborate with each other to locate boundary nodes and collection of information. Moreover, these algorithms accurately detect the location of the object with reduced energy consumption and data transmission and also extends the life time of the network [32], [33].
Cluster based continuous monitoring mechanisms and applications for efficient data aggregation provide dynamic event driven cluster formation. Yuan et al. introduces an energy efficient adaptive overlapping clustering method (EEAOC) for continuous monitoring. Using same overlapping structure of clusters, two adjacent sensor nodes can be merged into the same cluster for data aggregation and transmission operations. To overcome the quality of service requirements in the continuous monitoring applications also provide hybrid communication procedure by swapping time and event driven mechanisms. The hybrid communication procedure is operated to enhance the accuracy of detection and efficiency of energy [34].
Mahmuda et al. presents energy efficient tracking and localization mechanisms in wireless sensor networks to enhance the network life time. At the boundary areas, an energy efficient cluttering algorithm and Gaussian adaptive resonance theory is presented to aggregate patterns of sensor nodes, clustering patterns on the basis of sensing ranges and efficiently arranging the revived information. To modify instance motion patterns, it allows dynamically creation, learning and updating of the cluster. At the boundary of static clusters incremental clusters are formulate for continuous object tracking within a whole network It also provides accurate localization of dynamic objects using trilateration method [16].

B. CONTINUOUS OBJECT TRACKING AND BOUNDARY DETECTION SCHEMES
In cluster based tracking method network is divided into clusters to support data processing and collaborative communication between sensors nodes and Cluster Head (CH). Different clustering schemes are used such as static cluster, dynamic cluster and hybrid cluster. Xiang Ji et al. proposed a dynamic cluster structure for detecting and tracking (DCSODT) an accurate object boundary. In this approach clustering technique is proposed to minimize the communication overhead. In each cluster, boundary nodes are structured dynamically. The CH gathers information of nearby boundary from other boundary nodes of the cluster and report back to the sink. During boundary change, each CH update its members. This boundary movement speed effects the detection and tracking method. Main drawback of this scheme is energy consumption overhead that is generated when every CH directly/ultimately sends the data to the BS [35].
Chang WR et al. proposed a hybrid cluster with novel Continuous Object Detection and Tracking Algorithm (CODA). CODA uses static/dynamic (hybrid) clustering technique continuous objects detection and tracking, i.e. oil spill, and wild fire. In this mechanism each sensor node monitors and tracks the object boundary that moves within sensing region. Initially, static clusters are configured in the network and sensors nodes are deployed into these clusters. In each cluster, boundary nodes that detected the target they send their sensory data to its own CH through one hop selection method. When the target boundary is detected within each static cluster then each CH organized the BNs with its own cluster and forms a dynamic cluster. Each CH computes boundary values and sends aggregation data to sink. After that, SN computes the boundary for whole object in the region. In this method, cluster construction and maintenance overhead consumes energy [36]. It does not support in emergency situation by the congested data at SN. One hop selection method for reporting the boundary information to the CH also causes energy overhead. It is not suitable for concave polygon shape [37].
Taj Rehman, et al. presented a scheme CDCAPC consistent data collection and assortment in the progression of continuous objects in IoT. In this scheme, congestion diminish, minimizing the data injecting rate and maximum throughput problem is tackled by taking the different link capacity, congested BN selection and residual power of nodes. Congestion occurs when sensor nodes transmit the data packets to the same target node and their speed of data injecting is greater than the outgoing capacity of the link and processing of the target node. It causes severe damage in critical situation. To tackle this situation a hotspot is formed for consistent data gathering and transmission. For minimization of buffered based packet drop ratio and congestion PCCS preliminary congestion control stage algorithm is used that select the parent node and serve the one data flow among multiple data flows. Parent node selects the child node which has lower data transfer rate and high priority. If the buffer overflowing is still on hand, RBNIC representative boundary nodes identification and congestion control algorithm executes for selecting the uncongested parent node for congestion mitigation. For transferring the data to the parent node few RNs are used. It is obvious that there is no need of congested nodes to neighboring with regular nodes and transmit data to its parent node. After that, hotspot is eliminated and every node checked the next hop by its flag value and calculates the link capacity to transmit the data. This scheme does not provide selection mechanism for boundary nodes only assumptions are created for BNs. Heavy data load may cause energy consumption [38].
In Energy efficient and accurate tracking and detection of continuous objects in WSNs (EEATDC), T. Rehman et al. proposed a two-level boundary localization and detection technique for duty-cycled wireless sensor networks to achieve boundary accuracy through reduction in energy consumption. In duty cycle deployment, a few nodes are active and remaining nodes are in sleeping state which cause energy consumption and reduction in data traffic overhead. When an event occurs, planarized graphs are used to detect the object's coarse boundary and refine the boundary face construction. The sensors around boundary face nodes use four spatial interpolation methods (SIMs) to refine the boundary without awaking the sleeping sensors. Therefore, the candidate inner and outer boundary nodes are selected based on their estimated data to awake and report their data to the SN. The boundary area can be refined considerably by iteration of this refinement procedure. Planarization granularity becomes a problem in boundary nodes [39] when the short sensing range is used in fast object and high dense deployment. There are more chances of sensor nodes to get close to the real detected boundary to send more accurate information to the BS and less energy consumption. However, in some cases long sensing range is used when tracking a slow object and sparse deployment for accurate boundary line but it will cost more energy. Therefore, the short sensing range is better. In case of a static CH, dynamic clustering formation minimizes redundant data transmission [37]. The dynamic cluster formation reduces the number of messages forward to the sink. Therefore, the overall communication cost in terms of energy consumption is reduced [40].
In WSN, continuous nature objects have irregular attributes of expansion and contraction and these attributes cause detection and tracking more challenging. Sensor nodes provide the location of the continuous object and SN estimate the boundary of object using the received information and accuracy of the information based on the boundary node and failure of boundary node affects the boundary detection procedure. Therefore, Sajida et al. presents an efficient failureprone object detection scheme that uses spatial and temporal attributes of the sensor nodes to detects and recover the failure occur at the boundary nodes without compromising the accuracy of boundary detection procedure. It detects the change and assigns weight to the boundary nodes and restricts the boundary node selection procedure on one and two hop neighbors to achieve optimality. For leader node selection back-off-timer is set at each boundary node as shown in equation (1). BN 1d denotes normal boundary nodes, w 1 and w 2 denotes weak and normal boundary nodes, residual energy represents by E rd and stronger bins with BBN d to select leader node on the basis of maximum energy and a strong node.
For efficiency, leader node is selected on the basis of higher number of neighboring boundary nodes and residual energy to aggregate data from the other boundary nodes and transmits towards the sink [41]. Hyun et al. presents a subset selection based algorithm to identify the mobile object. A data aggregation based algorithm introduces for reducing the amount of report messages and sending information to representative nodes. Quadratic polynomial interpolation algorithm for detecting boundary nodes and restore the shape of the boundary [42]. A mechanism utilize hybrid network introduces by Jianming et al. to fill up sensing holes and selecting locations for the mobile sensing nodes. Before Network initialization, voronoi diagram utilize for hole detection [43], [44]. Static sensor nodes detect the value toxic air and variation detected by the presented mechanism and estimating the area where variation values are high. Optimum location selection mechanism detecting the variation points for boundary detection and sensor holes also fill up by the mobile sensors. Sensing holes selected on the basis of spatial and variation factors and maintain target list for mobile nodes [45].
For removal of redundancy in space and time domain, a time domain based adaptive sampling was presented. To overcome the designing issue of adaptive sampling in continuous object tracking a new method is presented to visualize the whole sensing field into number of small cells like pixels on visual screens by adopting sensor grid in static clustering based wireless sensor network architecture. To reduce redundancy information of boundary like boundary node, control message and report packets, a picture based image of diffusing objects consider for selecting boundary in the space domain [46].

C. CONTINUOUS OBJECT BOUNDARY DETECTION SCHEMES
The energy efficient continuous boundary monitoring (COBOM) approach was presented for energy-efficient boundary detection. The sensor node broadcast its ID to its one-hop neighbor and matched its current reading. If the current reading is not matched to previous reading, then it saves the information in node's BN-array and then that node becomes a BN. A small number of RNs are chosen for energy efficiency. It is challenging to determine the location for RN's neighbors as BN-Array contains the detection data instead of IDs. The other issue with COBOM is that the BNs and RNs are formed on both sides of the Boundary. It results in yielding more number of BNs and RNs; thus, increasing the communication cost [47].
Y. Zhang et al. offered a scheme (BRDCO) for detecting the boundary region of continuous objects using fog computing by deploying some mobile sensors in the sparse network. By the help of planar algorithms network is initially divided into sub regions and builds a routing map for data transmission. Static nodes are deployed by applying fixed sensing range in sparse network which results in sensing holes' existence. These nodes will transmit their positioning and sensing data to BS by multi hop communication within the region. After analyzing the data, spatial interpolation algorithm (IDW) is applied to calculate the sensory data and predicate the object boundary line. For removing sensing holes, mobile nodes are applied in order to pass through these boundary locations and collect the data for predicting more accurate boundary. Mobility of nodes reduce energy consumption by applying ACO an ant colony optimization algorithm for path optimization of mobile sensors nodes. Transmission overhead occurs by reporting to the sink. Wide adoption of mobile sensors is costly and these are not well suited option for gathering data in special environments, i.e. mountains and uneven harsh places [48].
A technique was presented for accurate and energyefficient boundary detection of continuous objects in duty-cycled WSN (EEBDDC). Based on densely deployed activated sensors nodes, a two-level boundary face detection method is proposed. Initially set the certain nodes in duty cycle(active) conditions to monitor the situation. These active sensors recognized the potential events when occurs and also detected the boundary faces of continuous objects by adopting four types of planarization algorithms Gabriel graph (GG), (RNG) relative neighborhood graph, Yao graph (YG) and (LDelk) k-localized Delaunay graph. Most of the sensors nodes are in sleep state that because coarse boundary faces somehow. To remove this coarse boundary face, four spatial interpolation methods (SIMs) are adopted for estimating the data of boundary nodes. Based on the sensory data of some nodes, more appropriate boundary face nodes are selected as candidate BNs. These nodes are awakened and route the data to the sink. As a result, sizes of boundary faces are minimized. Thereafter, boundary area will be refined considerably by iteration of this refinement procedure. Assumptions of the Planarization algorithms are not relevant to all cases. Planarization granularity becomes a problem in boundary nodes [49].
Lei et al. presented a scheme for detection and visualization of the dangerous zone in wireless sensor networks. It uses five different types of planarization algorithms to plenarize a WSN and detect the dangerous area of the by partition into inner and outer boundary area for the leakage of gases. Adopt planarization to obtain different topologies and also analyze the impact of these planarization algorithms on the detection of dangerous zone of the gas while considering the node failure scenario. Although presented scheme discussed briefly and analyzed for different case. However, no proper contribution from presenting scheme for highlighted cases on which different existing schemes are analyzed [50].

III. ANALYSIS OF SCHEMES
In this section, we conducted a comparative analysis of schemes and adjust different schemes data in tabular form. Analysis conducted according to their main idea, method, limitations and strong points of multiple schemes are given in Table 2. We conducted an analytical review of these schemes as elaborated in tabular form as shown in table 2. This table comprises of continuous object tracking and boundary detection schemes. These schemes are divided into three sections according to the taxonomy. The schemes including [28], [30]- [32], [34] and [16] provide energy efficient mechanisms for continuous object detection. In [30]- [32], [34], communication overhead is reduced. In [31], a hybrid algorithm based on COBOM and DEMOCO reduces the cost and enhances the accuracy. The cluster based schemes [16] and [34] involve efficient mechanisms for object tracking. In the similar vein, an efficient mechanism [28] uses tree based approach. In [35], [37], [39] and [46], communication cost is reduced and life time is enhanced. Moreover, energy consumption was improved in [39]. In [41], [42], [45] and [46], efficient object and boundary detection mechanisms were presented whereas fault detection and recovery mechanism were explored n [41]. The redundancy removal mechanism was formulate in [46] to reduce the communication cost of the network while selecting the boundary of the object.
To avoid congestion, a mechanism was presented in [38] to reduce congestion for transmission and packet drop ratio. In [47], [48], energy efficient and accurate boundary detection mechanisms were presented. In [47], the packet size was also reduced to diminishing the transmission cost. In [49], four planarization algorithms graph are adopted for identifying boundary faces. Moreover, four spatial interpolation methods are used for estimating the sensory data of boundary sensor nodes. In [50], planarization based five algorithms were adopted to detect and visualize the impact of gas in dangerous area boundary with fault tolerate. Analysis explores that energy efficient object tracking and boundary detection demands reduced communication overhead, data redundancy and packet drop ratio. Accuracy is also a significant factor for object and boundary detection in emergency scenarios.

IV. OPEN RESEARCH CHELLANGES
The existing literature covers a variety of solutions for continuous object tracking. However, new challenges arise in different application scenarios. We explored a number of open research challenges as follows.

A. SENSING RANGE AND ENERGY CONSUMPTION
Sensing range affects the performance of network and energy efficiency during continuous object tracking. It involves a large amount of communication among various sensor nodes [39], [51]. There is a tradeoff between sensing range and energy consumption [61]. Large sensing range can reduce packet drop ratios and reducing communication overhead but it requires more energy for each message transmission. It is quite challenging to manage it efficiently to enhance network lifetime [52], [53].
The boundary nodes interact with each other and also report back to the BS for tracking the continuous object by consuming additional energy [54], [59], [60]. Therefore, size of report message is reduced for energy efficiency in [30] and if only those nodes send data to CH which have new boundary information. As a result, it conserves the total energy and prolongs the overall network lifetime [37], [62]. It is quite challenging to choose short range or long sensing range. If there are small numbers of nodes in the sparse network to report back to the BS, then it would require larger transmission ranges to make communication possible [58]. In dense network, for overlapped sensing areas it is important to consider more short sensing range of sensor nodes. The sensor node should adjust its sensing range according to the specific conditions so that it saves energy and monitor the object for longer time [47]. In [55] and [56], sensor nodes  were organized in sets for energy constraints because extensive amount of message exchange between nodes consumed more energy and communication overhead occurs. It aimed to cover large set of targets sensor nodes. In [57], the main focus is for known location targets monitoring where less power is consumed by keeping other nodes away from the communication.

B. ACCURACY OF BOUNDARY DETECTION
In continuous object tracking, accurate boundary detection of object is more challenging aspect [63], [64] which demands proper nodes deployment [65]. Continuous objects have a tendency to change shapes as well as to blowout under the wind pressure, therefore, accurate and timely tracking of boundary movement of such objects is a challenging issue [37]. Toxic gas leakage and diffusion are difficult processes that cause severe damages and air pollution. It is quite critical to timely and accurately identify the source and spreading direction [47], [30]. Although, the schemes [41], [49], [39] achieve accurate boundary detection through duty cycle sensor node in dense network. However, accuracy can be further improved [66]. The IoT enabled applications track the toxic gas leakage in industrial area [67], [68], oil spills and fire detection. It demands the accurate boundary detection for dependable solutions [48]. It opens a new set of challenges that must be resolved by the researchers through innovative solutions.

C. BOUNDARY FACE AREA LOCALIZATION
The boundary face area detection is quite challenging while achieving accuracy and reducing the energy consumption [49], [73]. In [39], the energy consumption is reduced by sharing the messages to the nodes that are located on critical positions near boundary face. It is also quite challenging to discover more number of active nodes that can play as backup role for the critical nodes in boundary face area. It is quite challenging task to accurately predict the continuous object while it is spreading in multiple directions with different speeds. Thus issue is solved through localization of boundary face area. The researchers should also present the schemes that predict the chances of spread towards nearby critical locations where precious and valuable records or items are placed. In case of fire, there may be the nearby petroleum reservoirs or ammunition stores that must be protected by identifying the boundary face and its spread with a certain speed.

D. TRANSMISSION OVERHEAD
It is quite challenging to reduce the communication cost while ensuring the accuracy and timeliness for continuous objects tracking and detection which involves large number of sensors [69], [72]. It becomes more challenging when continuous object disperses in an extensive physical area or move with a high speed [71]. Communication overhead is minimized through duty cycle sensor deployment in dense network to improve boundary face area [37], [39], [49]. In emergency situation, data transmission is extremely challenging factor to detect the event boundary and effective utilization of network link capacity. It requires massive communication and transmission overhead when all the sensor nodes report the data to the sink. It also causes congestion which results in severe packet loss and collision. To address this challenge, the scheme [38] minimized the communication cost, energy consumption and utilize the maximum throughput. To further improve it, the fog nodes require high computational capacity to process and analyze the data from sensors [70]. It is quite challenging to achieve accurate boundary with minimum transmission overhead in real-time scenarios [48].

E. CONGESTION AND DATA LOSS
It is a challenging issue in dense deployment where congestion occurs due to excessive communication in a specific time bracket [74]. It results in data loss even in emergency scenarios [49]. To address this challenge, few active nodes are used to monitor the continuous object and minimize the congestion and data traffic, rest of the nodes are in sleeping state [39], [49]. It reduces the size of boundary faces and the data traffic is also minimized. To reduce congestion, the researchers should focus on improving the data packet loss, end-to-end and hop-by-hop delay, high precedence data delivery as in [38], [72].

F. REDUCTION OF ACTIVE NODES
A continuous object usually spread in large areas which involve a large number of active sensor nodes to detect and track it. Massive communication is involved when all the incident nodes transmit their sensory data and position information to BS. Generally, energy consumption is reduced by decreasing the number of boundary nodes and reducing the reporting information size [41]. To address this challenge, the continuous objects should be monitored by minimizing the communication overhead with fewer number of active nodes selection [39], [49]. It is also challenging to identify the trajectory of move for continuous object which involves taking data from a large number of active senor node. It opens a new horizon for the researchers to present dependable solutions that involve less number of active nodes and ensure the reliable communication without missing any relevant data from critical region.

V. CONCLUSION
IoT empowered applications are quite beneficial for the tracking and monitoring of continuous objects including toxic gas, oil spills and fire. In this paper, we have investigated the existing and state of the art work on continuous object tracking and boundary detection. We presented a taxonomy to arrange the literature for both object and boundary detection mechanism. Moreover, FoG assisted structural design are also explored for boundary detection. We performed the analysis for the schemes in literature. It has been identified that mostly object tracking and boundary detection techniques aim to diminish energy consumption, congestion and transmission overhead. Moreover, duty cycle mechanism of sensors is used for minimizing the energy consumption to prolong the network lifetime. Finally, we identified the open research challenges that must be solved to provide the dependable solutions. In future, we shall analyze the deployments where drones can extract the data from sensor nodes and CH as well during continuous object racking.
N. Z. JHANJHI is currently working as the Director of the Center for Smart Society 5.0 [CSS5], an Associate Professor, and the Cluster Head of Cybersecurity Cluster at the School of Computer Science and Engineering, Faculty of Innovation and Technology, Taylor's University, Malaysia. He is supervising a great number of postgraduate students mainly in the cybersecurity for data science. Cybersecurity Research Cluster has extensive research collaboration globally with several institutions and professionals. He has high indexed publications in WoS/ISI/SCI/Scopus and his collective research impact factor is more than 300 points as of first half of 2021. He has international patents on his account and edited/authored more than 29 research books published by world-class publishers. He has a great experience of supervising and co-supervising postgraduate students, an ample number of Ph.D. and master's students graduated under his supervision. He is an External Ph.D./Master's Thesis Examiner/an Evaluator for several universities globally. He has completed more than 22 international funded research grants successfully. His research interests include cybersecurity, the IoT security, wireless security, data science, software engineering, and UAVs. He is a PC Member for several IEEE conferences worldwide. An active reviewer for a series of Q1 journals, has been awarded globally as a top 1% reviewer by Publons (Web of Science). He has been awarded as an ''Outstanding Associate Editor'' by IEEE ACCESS for the year 2020. He has served as a keynote speaker for several international conferences, presented several webinars worldwide, and chaired international conference sessions. He is a guest editor for the reputed indexed journals. He is an Associate Editor and an Editorial Assistant of the board for several reputable journals including, IEEE ACCESS journal and PeerJ Computer Science.
MAMOONA HUMAYUN received the Ph.D. degree in computer architecture from Harbin Institute of Technology, China. She has 12 years of teaching and administrative experience internationally. She is an active reviewer for a series of journals. She has supervised various master's and Ph.D. thesis. Her research interests include global software development, requirement engineering, knowledge management, cybersecurity, and wireless sensor networks.
THAMER A. TABBAKH received the master's and Ph.D. degrees from CREOL, UCF, which is consider as top ten optics and photonics college in the USA. He is an Assistant Professor of electrical engineering and optics and photonics. He is currently the Head of Labs and Equipment Administration at KACST. He is currently working at KACST, which is considered as the largest research institution in Saudi Arabia.
ZAHRAH A. ALMUSAYLIM received the B.S. and M.S. degrees in computer science from the College of King Faisal University (KFU), Saudi Arabia. She is a Researcher at King Abdulaziz City for Science and Technology (KACST), Saudi Arabia. She is currently the Head of Digital Transformation of Labs Administration at KACST. Her research interests include the Internet of Things, cloud computing, the IoT privacy and security, wireless sensor networks, security of RPL networks, cloud computing security, network security, mobile computing, context-aware computing, machine learning, and web and mobile applications programming.