By Topic

Distributed Smart Cameras, 2009. ICDSC 2009. Third ACM/IEEE International Conference on

Date Aug. 30 2009-Sept. 2 2009

Filter Results

Displaying Results 1 - 25 of 80
  • Target detection and counting using a progressive certainty map in distributed visual sensor networks

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1021 KB) |  | HTML iconHTML  

    Visual sensor networks (VSNs) merge computer vision, image processing and wireless sensor network disciplines to solve problems in multi-camera applications by providing valuable information through distributed sensing and collaborative in-network processing. Collaboration in sensor networks is necessary not only to compensate for the processing, sensing, energy, and bandwidth limitations of each sensor node but also to improve the accuracy and robustness of the sensor network. Collaborative processing in VSNs is more challenging than in conventional scalar sensor networks (SSNs) because of two unique features of cameras, including the extremely higher data rate compared to that of scalar sensors and the directional sensing characteristics with limited field of view. In this paper, we study a challenging computer vision problem, target detection and counting in VSN environment. Traditionally, the problem is solved by counting the number of intersections of the backprojected 2D cones of each target. However, the existence of visual occlusion among targets would generate many false alarms. In this work, instead of resolving the uncertainty about target existence at the intersections, we identify and study the non-occupied areas in the cone and generate the so-called certainty map of non-existence of targets. This way, after fusing inputs from a set of sensor nodes, the unresolved regions on the certainty map would be the location of target. This paper focuses on the design of a light-weight, energy-efficient, and robust solution where not only each camera node transmits a very limited amount of data but that a limited number of camera nodes is used. We propose a dynamic itinerary for certainty map integration where the entire map is progressively clarified from sensor to sensor. When the confidence of the certainty map is satisfied, a geometric counting algorithm is applied to find the estimated number of targets. In the conducted experiments using real data, the resu- lts of the proposed distributed and progressive method shows effectiveness in detection accuracy and energy and bandwidth efficiency. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Metric learning for semi-supervised clustering of Region Covariance Descriptors

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (684 KB) |  | HTML iconHTML  

    In this paper we extend distance metric learning to a new class of descriptors known as region covariance descriptors. Region covariances are becoming increasingly popular as features for object detection and classification over the past few years. Given a set of pairwise constraints by the user, we want to perform semi-supervised clustering of these descriptors aided by metric learning approaches. The covariance descriptors belong to the special class of symmetric positive definite (SPD) tensors, and current algorithms cannot deal with them directly without violating their positive definiteness. In our framework, the distance metric on the manifold of SPD matrices is represented as an L2 distance in a vector space, and a Mahalanobis-type distance metric is learnt in the new space, in order to improve the performance of semi-supervised clustering of region covariances. We present results from clustering of covariance descriptors representing different human images, from single and multiple camera views. This transformation from a set of positive definite tensors to a Euclidean space paves the way for the application of many other vector-space methods to this class of descriptors. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Bayesian formulation of image patch matching using cross-correlation

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (8429 KB) |  | HTML iconHTML  

    A classical solution for matching two image patches is to use the cross-correlation coefficient. This works well if there is a lot of structure within the patches, but not so well if the patches are close to uniform. This means that some patches are matched with more confidence than others. By estimating this uncertainty more weight can be put on the confident matches than those that are more uncertain. In this paper we present a system that can learn the distribution of the correlation coefficient from a video sequence of an empty scene. No manual annotation of the video is needed. Two distributions functions are learned for two different cases: i) the correlation between an estimated background image and the current frame showing that background and ii) the correlation between an estimated background image and an unrelated patch. Using these two distributions the patch matching problem is formulated as a binary classification problem and the probability of two patches matching is derived. The model depends on the signal to noise ratio. The noise level is reasonably invariant over time, while the signal level, represented by the amount of structure in the patch or it's spatial variance, has to be measured for every frame. A common application where this is useful is feature point matching between different images. Another application is background/foreground segmentation. In this paper it is shown how the theory can be used to implement a very fast background/foreground segmentation by transforming the calculations to the DCT-domain and processing a motion JPEG stream without uncompressing it. This allows the algorithm to be embedded on a 150 MHz ARM based network camera. It is also suggested to use recursive quantile estimation to estimate the background model. This gives very accurate background models even if there is a lot of foreground present during the initialisation of the model. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Efficient approximate foreground detection for low-resource devices

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (919 KB) |  | HTML iconHTML  

    A broad range of very powerful foreground detection methods exist because this is an essential step in many computer vision algorithms. However, because of memory and computational constraints, simple static background subtraction is very often the technique that is used in practice on a platform with limited resources such as a smart camera. In this paper we propose to apply more powerful techniques on a reduced scan line version of the captured image to construct an approximation of the actual foreground without over burdening the smart camera. We show that the performance of static background subtraction quickly drops outside of a controlled laboratory environment, and that this is not the case for the proposed method because of its ability to update its background model. Furthermore we provide a comparison with foreground detection on a subsampled version of the captured image. We show that with the proposed foreground approximation higher true positive rates can be achieved. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Human interaction analysis based on walking pattern transitions

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (6814 KB) |  | HTML iconHTML  

    We propose a method that analyzes interaction between pedestrians based on their trajectories obtained using sensors such as cameras. Our objective is to understand the mutual relationship between pedestrians and to detect anomalous events in a video sequence. Under such situations, we can observe the interaction between a pair of pedestrians. This paper proposes a set of features that measures the interaction between pedestrians. We assume that a person is likely to change his/her walking patterns when he/she has been influenced by another person. Based on this assumption, the proposed method first extracts the transition points of a walking pattern from trajectories of two pedestrians and then measures the strength of the influence using the temporal and spatial closeness between them. Finally, experimental results obtained from actual videos demonstrate the method's effectiveness in understating mutual relationships and detecting anomalous events. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Multi-camera tracking on a graph using Markov chain Monte Carlo

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (816 KB) |  | HTML iconHTML  

    Wide-area surveillance requires a system of multiple cameras that are sparsely distributed without overlapping fields of view. Tracking objects in such a setting is challenging because blind gaps between disjoint camera views cannot ensure spatial, temporal, and visual continuity in successive observations. We propose an association algorithm for tracking an unknown number of objects with sparsely distributed uncalibrated cameras. To model traffic patterns in a monitored environment, we exploit the statistics on overall traffic and the probabilistic dependence of a path in one camera view on the previous path in another camera view. The dependency and the frequency of allowable paths are represented in a graph model. Without using a high-order transition model, the proposed graph disambiguates traffic patterns and generalizes traffic constraints in motorway and indoor scenarios. Based on the graph model, we derive a posterior probability of underlying paths, given a set of observations. The posterior evaluates not only the plausibility of individual paths but also the hypothesized number of paths with respect to traffic statistics of the environment. To find the maximum a posteriori, we use Markov chain Monte Carlo (MCMC). In contrast to other MCMC-based tracking methods, the proposed MCMC sampling requires neither additional cost to compute an initial sample nor information about the number of objects passing through the environment. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Automatic camera selection for activity monitoring in a multi-camera system for tennis

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (514 KB) |  | HTML iconHTML  

    In professional tennis training matches, the coach needs to be able to view play from the most appropriate angle in order to monitor players' activities. In this paper, we describe and evaluate a system for automatic camera selection from a network of synchronised cameras within a tennis sporting arena. This work combines synchronised video streams from multiple cameras into a single summary video suitable for critical review by both tennis players and coaches. Using an overhead camera view, our system automatically determines the 2D tennis-court calibration resulting in a mapping that relates a player's position in the overhead camera to their position and size in another camera view in the network. This allows the system to determine the appearance of a player in each of the other cameras and thereby choose the best view for each player via a novel technique. The video summaries are evaluated in end-user studies and shown to provide an efficient means of multi-stream visualisation for tennis player activity monitoring. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Distributed and lightweight multi-camera human activity classification

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (4452 KB) |  | HTML iconHTML  

    We propose a human activity classification algorithm that has a distributed and lightweight implementation appropriate for wireless camera networks. With input from multiple cameras, our algorithm achieves invariance to the orientation of the actor and to the camera viewpoint. We conceptually describe how the algorithm can be implemented on a distributed architecture, obviating the need for centralized processing of the entire multi-camera data. The lightweight implementation is made possible by the very affordable memory and communication bandwidth requirements of the algorithm. Notwithstanding its lightweight nature, the performance of the algorithm is comparable to that of the earlier multi-camera approaches that are based on computationally expensive 3D human model construction, silhouette matching using reprojected 2D views, and so on. Our algorithm is based on multi-view spatio-temporal histogram features obtained directly from acquired images; no background subtraction is required. Results are analyzed for two publicly available multi-camera multi-action datasets. The system's advantages relative to single camera techniques are also discussed. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Detection of composite events spanning multiple camera views with wireless embedded smart cameras

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (859 KB) |  | HTML iconHTML  

    With the introduction of battery-powered and embedded smart cameras, it has become viable to install many spatially-distributed cameras interconnected by wireless links. However, there are many problems that need to be solved to build scalable, battery-powered wireless smart-camera networks (Wi-SCaNs). These problems include the limited processing power, memory, energy and bandwidth. Limited resources necessitate light-weight algorithms to be implemented and run on the embedded cameras, and also careful choice of when and what data to transfer. We present a wireless embedded smart camera system, wherein each camera platform consists of a camera board and a wireless mote, and cameras communicate in a peer-to-peer manner over wireless links. Light-weight background subtraction and tracking algorithms are implemented and run on camera boards. Cameras exchange data to track objects consistently, and also to update locations of lost objects. Since frequent transfer of large-sized data requires more power and incurs more communication delay, transferring all captured frames to a server should be avoided. Another challenge is the limited local memory for storage in camera motes. Thus, instead of transferring or saving every frame or every trajectory, there should be a mechanism to detect events of interest. In the presented system, events of interest can be defined beforehand, and simpler events can be combined in a sequence to define semantically higher-level and composite events. Moreover, event scenarios can span multiple camera views, which make the definition of more complex events possible. Cameras communicate with each other about the portions of a scenario to detect an event that spans different camera views. We present examples of label transfer for consistent tracking, and of updating the location of occluded or lost objects from other cameras by wirelessly exchanging small-sized packets. We also show examples of detecting different composite and spatio-temporal - event scenarios spanning multiple camera views. All the processing is performed on the camera boards. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Unsupervised camera network structure estimation based on activity

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (479 KB) |  | HTML iconHTML  

    In this paper we consider the problem of unsupervised topology reconstruction in uncalibrated visual sensor networks. We assume that a number of video cameras observe a common scene from arbitrary and unknown locations, orientations and zoom levels, and show that the extrinsic and calibration matrices, fundamental and essential matrices, the homography matrix, and the physical configuration of the cameras with respect to each other can be estimated in an unsupervised manner. Our method relies on the similarity of activity patterns observed at various locations, and an unsupervised matching method based on these activity patterns. The proposed method works in cases with cameras having significantly different orientations and zoom levels, where many of the existing methods cannot be applied. We explain how to extend the method to a multicamera case where more than two cameras are involved. We present both qualitative and quantitative results of our estimates, and conclude that this method can be applied in wide area surveillance applications in which the deployed systems need to be flexible and scalable, and where calibration can be a major challenge. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Covariance descriptors on moving regions for human detection in very complex outdoor scenes

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (3184 KB) |  | HTML iconHTML  

    The detection of humans in very complex scenes can be very challenging, due to the performance degradation of classical motion detection and tracking approaches. An alternative approach is the detection of human-like patterns over the whole image. The present paper follows this line by extending Tuzel et al.'s technique based on covariance descriptors and LogitBoost algorithm applied over Riemannian manifolds. Our proposal represents a significant extension of it by: (a) exploiting motion information to focus the attention over areas where motion is present or was present in the recent past; (b) enriching the human classifier by additional, dedicated cascades trained on positive and negative samples taken from the specific scene; (c) using a rough estimation of the scene perspective, to reduce false detections and improve system performance. This approach is suitable in multi-camera scenarios, since the monolithic block for human-detection remains the same for the whole system, whereas the parameter tuning and set-up of the three proposed extensions (the only camera-dependent parts of the system), are automatically computed for each camera. The approach has been tested on a construction working site where complexity and dynamics are very high, making human detection a real challenge. The experimental results demonstrate the improvements achieved by the proposed approach. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A pervasive smart camera network architecture applied for multi-camera object classification

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (627 KB) |  | HTML iconHTML  

    Visual sensor networks are an emerging research area with the goal of using cameras as pervasive and affordable sensing and processing devices. This paper presents a pervasive smart camera platform which is built from off-the-shelf hardware and software components. The hardware platform is comprised of an OMAP 3530 processor, 128 MB RAM and various interfaces for connecting sensors and peripherals. A dual-radio wireless network allows to trade communication performance for power consumption. The software architecture is built upon standard Linux and supports dataflow oriented application development by dynamically instantiating and connecting functions blocks. Data is transferred between blocks via shared memory for high throughput. We present a performance evaluation of our smart camera platform as well as a multi-camera object classification system to demonstrate the capabilities and applicability of our approach. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A framework for determining overlap in large scale networks

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1107 KB) |  | HTML iconHTML  

    This paper presents a novel framework designed for calculating the topology of overlapping cameras in large surveillance systems. Such a framework is a key enabler for efficient network-wide surveillance, e.g. inter-camera tracking, especially in large surveillance networks. The framework presented can be adapted to utilise numerous contradiction and correlation approaches to identify overlapping portions of camera views using activity within the system. It can also utilise a various arbitrary occupancy cells which can be used to adjust both the memory requirements and accuracy of the topology generated. The framework is evaluated for its memory usage, processing speed and the accuracy of its overlap topology on a 26 camera dataset using various approaches. A further examination of memory requirements and processing speed on a larger 200 camera network is also presented. The results demonstrate that the framework significantly reduces memory requirements and improves execution speed whilst producing useful topologies from a large surveillance system at real-time speeds. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Color Brightness Transfer Function evaluation for non overlapping multi camera tracking

    Page(s): 1 - 6
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (464 KB) |  | HTML iconHTML  

    People Tracking in multiple cameras is of great interest for wide area video surveillance systems. Multi-camera tracking with non-overlapping fields of view (FOV) involves the tracking of people in the blind region and their correspondence matching across cameras. We consider these problems in this paper. We propose a multi camera architecture for wide area surveillance and a real time people tracking algorithm across non overlapping cameras. We compared different methods to evaluate the color Brightness Transfer Function (BTF) between non overlapping cameras. These approaches are based on a testing phase during which the color histogram mapping, between pairs of images of the same object observed in the different field of views, is carried out. The experimental results compare two different transfer functions and demonstrate their limits in people association when a new person enters in one camera FOV. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Dependable integrated surveillance systems for the physical security of metro railways

    Page(s): 1 - 7
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (394 KB) |  | HTML iconHTML  

    Rail-based mass transit systems are vulnerable to many criminal acts, ranging from vandalism to terrorism. In this paper, we present the architecture, the main functionalities and the dependability related issues of a security system specifically tailored to metro railways. Heterogeneous intrusion detection, access control, intelligent video-surveillance and sound detection devices are integrated in a cohesive Security Management System (SMS). In case of emergencies, the procedural actions required to the operators involved are orchestrated by the SMS. Redundancy both in sensor dislocation and hardware apparels (e.g. by local or geographical clustering) improve detection reliability, through alarm correlation, and overall system resiliency against both random and malicious threats. Video-analytics is essential, since a small number of operators would be unable to visually control a large number of cameras. Therefore, the visualization of video streams is activated automatically when an alarm is generated by smart-cameras or other sensors, according to an event-driven approach. The system is able to protect stations (accesses, technical rooms, platforms, etc.), tunnels (portals, ventilation shafts, etc.), trains and depots. Presently, the system is being installed in the Metrocampania underground regional railway. To the best of our knowledge, this is the first subway security system featuring artificial intelligence algorithms both for video and audio surveillance. The security system is highly heterogeneous in terms not only of detection technologies but also of embedded computing power and communication facilities. In fact, sensors can differ in their inner hardware-software architecture and thus in the capacity of providing information security and dependability. The focus of this paper is on the development of novel solutions to achieve a measurable level of dependability for the security system in order to fulfill the requirements of the specific application. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Performance evaluation of two state of the art DVC codecs

    Page(s): 1 - 4
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (249 KB) |  | HTML iconHTML  

    The performance of existing DVC codecs is still lacking relative to that of H.264 and work is being carried out in order to close this gap. The authors of this paper have been and still are involved in the development of two DVC codecs respectively, the performance of which is compared herein. The aim is to identify strengths and weaknesses of the two codecs that can be exploited/addressed in order to improve the achieved performance relative to H.264. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Efficient topology calibration and object tracking with distributed pan tilt cameras

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1739 KB) |  | HTML iconHTML  

    We propose a method for calibrating the topology of distributed pan tilt cameras (i.e., the structure of routes among FOVs) and its probabilistic model, which is useful for multi-object tracking in a wide area. To observe objects as long and many as possible, pan tilt control is an important issue in automatic calibration as well as in tracking. If only one object is observed by a camera and its neighboring cameras, the camera should point towards this object both in the calibration and tracking periods. However, if there are multiple objects, in the calibration period, the camera should be controlled towards an object that goes through an unreliable route in which a sufficient number of object detection results have not been observed. This control allows us to efficiently establish the reliable topology model. After the reliable topology model is established, on the other hand, the camera should be directed towards the route with the biggest possibility of object observation. We therefore propose a camera control framework based on the mixture of the reliability of the estimated routes and the probability of object observation. This framework is applicable both to camera calibration and object tracking by adjusting weight variables. Experiments demonstrate the efficiency of our camera control scheme for establishing the camera topology model and tracking objects as long as possible. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • An efficient system for vehicle tracking in multi-camera networks

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1088 KB) |  | HTML iconHTML  

    The recent deployment of very large-scale camera networks has led to a unique version of the tracking problem whose goal is to detect and track every vehicle within a large urban area. To address this problem we exploit constraints inherent in urban environments (i.e. while there are often many vehicles, they follow relatively consistent paths) to create novel visual processing tools that are highly efficient in detecting cars in a fixed scene and at connecting these detections into partial tracks.We derive extensions to a network flow based probabilistic data association model to connect these tracks between cameras. Our real time system is evaluated on a large set of ground-truthed traffic videos collected by a network of seven cameras in a dense urban scene. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Autonomous real-time surveillance system with distributed IP cameras

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (910 KB) |  | HTML iconHTML  

    An autonomous Internet Protocol (IP) camera based object tracking and behaviour identification system, capable of running in real-time on an embedded system with limited memory and processing power is presented in this paper. The main contribution of this work is the integration of processor intensive image processing algorithms on an embedded platform capable of running at real-time for monitoring the behaviour of pedestrians. The Algorithm Based Object Recognition and Tracking (ABORAT) system architecture presented here was developed on an Intel PXA270-based development board clocked at 520 MHz. The platform was connected to a commercial stationary IP-based camera in a remote monitoring station for intelligent image processing. The system is capable of detecting moving objects and their shadows in a complex environment with varying lighting intensity and moving foliage. Objects moving close to each other are also detected to extract their trajectories which are then fed into an unsupervised neural network for autonomous classification. The novel intelligent video system presented is also capable of performing simple analytic functions such as tracking and generating alerts when objects enter/leave regions or cross tripwires superimposed on live video by the operator. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Resolution mosaic-based Smart Camera for video surveillance

    Page(s): 1 - 7
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (743 KB) |  | HTML iconHTML  

    Video surveillance is one of the most data intensive applications. A typical video surveillance system consists of one or multiple video cameras, a central storage unit, and a central processing unit. At least two bottlenecks exist: First, the transmission capacity is limited, especially for raw data. Second, the central processing unit has to process the incoming data to give results in real time. Therefore, we propose an FPGA-based embedded camera system which performs all steps of image acquisition, region of interest extraction, generation of a multiresolution image, and image transmission. The proposed pipeline-based architecture allows a real time wavelet-based image segmentation and a detection of moving objects for surveillance purposes. The system is integrated in a single FPGA using external RAM as storage for images and for a Linux operating system which controls the data flow. With the pipeline concept and a Linux device driver it is possible to create a system for streaming the results of an image processing through a GbE interface. A real time processing is achieved. The camera transmits the captured images with 30 Mpixel/s, but the system is able to process 100 Mpixel/s. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Online video synthesis for removing occluding objects using multiple uncalibrated cameras via plane sweep algorithm

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1436 KB) |  | HTML iconHTML  

    We present an online rendering system which removes occluding objects in front of the objective scene from an input video using multiple videos taken with multiple cameras. To obtain geometrical relations between all cameras, we use projective grid space (PGS) defined by epipolar geometry between two basis cameras. Then we apply plane-sweep algorithm for generating depth image in the input camera. By excluding the area of occluding objects from the volume of the sweeping planes, we can generate the depthmap without the occluding objects. Using this depthmap, we can render the image without obstacles from all the multiple camera videos. Since we use graphics processing unit (GPU) for computation, we can achieve realtime online rendering using a normal spec PC and multiple USB cameras. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Recognizing activities from context and arm pose using finite state machines

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (772 KB) |  | HTML iconHTML  

    We present an activity-recognition system for assisted living applications and smart homes. While existing systems tend to rely on expensive computation of comparatively largedimension data sets, ours leverages information from a small number of fundamentally different sensor measurements that provide context information pertaining the person's location, and action information by observing the motion of the body and arms. Camera nodes are placed on the ceiling to track people in the environment, and place them in the context of a building map where areas and objects of interest are premarked. Additionally, a single inertial sensor node is placed on the subject's arm to infer arm pose, heading and motion frequency using an accelerometer, gyroscope and magnetometer. These four measurements are parsed using a lightweight hierarchy of finite state machines, yielding recognition rates with high precision and recall values (0.92 and 0.93, respectively). View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Tracking in sparse multi-camera setups using stereo vision

    Page(s): 1 - 6
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (279 KB) |  | HTML iconHTML  

    Tracking with multiple cameras with nonoverlapping fields of view is challenging due to the differences in appearance that objects typically have when seen from different cameras. In this paper we use a probabilistic approach to track people across multiple, sparsely distributed cameras, where an observation corresponds to a person walking through the field of view of a camera. Modelling appearance and spatio-temporal aspects probabilistically allows us to deal with the uncertainty but, to obtain good results, it is important to maximise the information content of the features we extract from the raw video images. Occlusions and ambiguities within an observation result in noise, thus making the inference less confident. In this paper, we propose to position stereo cameras on the ceiling, facing straight down, thus greatly reducing the possibility of occlusions. This positioning also leads to specific requirements of the algorithms for feature extraction, however. Here, we show that depth information can be used to solve ambiguities and extract meaningful features, resulting in significant improvements in tracking accuracy. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • 3D localization of projected objects for surveillance

    Page(s): 1 - 6
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (3353 KB) |  | HTML iconHTML  

    A localization method of projected objects in 3D work space is proposed in this paper, which uses a calibrated surveillance camera. The method determines a minimum bounding cylinder or box for a projected object in input image by using the intuition that lower boundary of the object lies on the ground. A base circle or rectangle corresponding to the bottom of the bounding solid is first estimated on the ground and then its size and height of the bounding solid are determined enough to enclose the 3D object corresponding to the projected object. This method can be applied to any free-shaped objects except humans. A particular method for projected human is also proposed, which estimates a base circle differently since the intuition may not be correct. The minimum bounding solid for a projected object is useful for determining its location in 3D surveillance space and estimating its volume roughly. Usefulness of the proposed methods is presented with experimental results on real moving objects. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Super-resolution based on blind deconvolution using similarity of power spectra

    Page(s): 1 - 7
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (218 KB) |  | HTML iconHTML  

    Generally, blind super-resolution with unknown blurs is treated as an optimization problem. This involves a cost function, composed of terms accounting for changes in image and point spread function (PSF), which usually undergoes regularization due to the ill-posedness of the problem. In this paper, we introduce a novel regularization term for the PSF such that the spectral change in the image caused by degradation is also included. This is based on the fact that the presence of PSF in images affects the frequency component concentration. This cost function is optimized with respect to the image and the PSF in an alternating manner. Experiment results show that the proposed method is effective based on an objective evaluation method and that its PSF estimation accuracy is competitive in comparison with the recently proposed parametric method. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.