Search

Understanding activity patterns in office environments is important in order to increase workers’ comfort and productivity. This paper proposes an automated system for discovering activity patterns of multiple persons in a work environment using a network of cheap low-resolution visual sensors (900 pixels). Firstly, the users’ locations are obtained from a robust people tracker based on recursive maximum likelihood principles. Secondly, based on the users’ mobility tracks, the high density positions are found using a bivariate kernel density estimation. Then, the hotspots are detected using a confidence region estimation. Thirdly, we analyze the individual’s tracks to find the starting and ending hotspots. The starting and ending hotspots form an observation sequence, where the user’s presence and absence are detected using three powerful Probabilistic Graphical Models (PGMs). We describe two approaches to identify the user’s status: a single model approach and a two-model mining approach. We evaluate both approaches on video sequences captured in a real work environment, where the persons’ daily routines are recorded over 5 months. We show how the second approach achieves a better performance than the first approach. Routines dominating the entire group’s activities are identified with a methodology based on the Latent Dirichlet Allocation topic model. We also detect routines which are characteristic of persons. More specifically, we perform various analysis to determine regions with high variations, which may correspond to specific events.

In this paper, we propose a novel extrinsic calibration method for camera networks using a sphere as the calibration object. First of all, we propose an easy and accurate method to estimate the 3D positions of the sphere center w.r.t. the local camera coordinate system. Then, we propose to use orthogonal procrustes analysis to pairwise estimate the initial camera relative extrinsic parameters based on the aforementioned estimation of 3D positions. Finally, an optimization routine is applied to jointly refine the extrinsic parameters for all cameras. Compared to existing sphere-based 3D position estimators which need to trace and analyse the outline of the sphere projection in the image, the proposed method requires only very simple image processing: estimating the area and the center of mass of the sphere projection. Our results demonstrate that we can get a more accurate estimate of the extrinsic parameters compared to other sphere-based methods. While existing state-of-the-art calibration methods use point like features and epipolar geometry, the proposed method uses the sphere-based 3D position estimate. This results in simpler computations and a more flexible and accurate calibration method. Experimental results show that the proposed approach is accurate, robust, flexible and easy to use.

Nowadays, an objective in visual communication is to send and store images of faces at a low bit rate, such that the faces are still recognizable and that the compression does not prevent remote face analysis. We present a novel segmented face approximation algorithm. Greyscale face images are segmented into meaningful surface segments with an adaptive region growing algorithm based on low-degree polynomial fitting. The algorithm uses a new adaptive thresholding technique with the \(L_\infty \) fitting cost as a segmentation criterion. The polynomial degree and the fitting error are automatically adapted during the region growing process. The main novelty is that the algorithm detects outliers and edges, distinguishes between strong and smooth intensity transitions, and finds surface segments that are bent in a certain way, such as flat, planar, convex, concave or saddle patches. As a result, the surface segments correspond to facial features, and the contours separating the surface segments coincide with real image face edges. Moreover, the curvature-based surface shape information facilitates many tasks in automated face analysis, demonstrated in this paper by face verification performed on the polynomial representation. The polynomial representation provides good approximation of facial features, while preserving all the necessary details of the face in the reconstructed image. When compared with different compression methods, we achieve higher compression ratios and better recognizable faces at low bit rates. This is confirmed by correct identification percentages obtained by face recognition algorithms on the compressed data.

Organizations that are constantly hiring new employees may run out of office space. Also, it is expensive to furnish and open new offices. Shared office space provides a cost-effective solution for big companies and startups. The main focus of our research is on understanding behavior patterns in work environments to increase workers' comfort and productivity. Specifically, the goal of this work is to detect habitual patterns of absence from a duty or obligation in order to provide a more efficient use of office space. In this paper, we propose to use a network of cheap low-resolution visual sensors (30 × 30 pixels) for long-term absenteeism detection of multiple persons in a work environment. Firstly, the users' locations are obtained from a robust people tracker based on recursive maximum likelihood principles. Secondly, based on the users' mobility tracks, an occupancy map with the hotspot locations is detected. Finally, we propose an algorithm for detecting the absence patterns of each person. We evaluate our method on video sequences captured in a real work environment, where the persons' daily routines are recorded over five months. The results show that our approach of detecting the absence patterns achieves an accuracy of 97.70% in comparison to ground truth.

In this paper, a novel approach for human gesture classification on skeletal data is proposed for the application of exergaming in physiotherapy. Unlike existing methods, we propose to use a general classifier like Random Forests to recognize dynamic gestures. The temporal dimension is handled afterwards by majority voting in a sliding window over the consecutive predictions of the classifier. The gestures can have partially similar postures, such that the classifier will decide on the dissimilar postures. This brute-force classification strategy is permitted, because dynamic human gestures show sufficient dissimilar postures. Online continuous human gesture recognition can classify dynamic gestures in an early stage, which is a crucial advantage when controlling a game by automatic gesture recognition. Also, ground truth can be easily obtained, since all postures in a gesture get the same label, without any discretization into consecutive postures. This way, new gestures can be easily added, which is advantageous in adaptive game development. We evaluate our strategy by a leave-one-subject-out cross-validation on a self-captured stealth game gesture dataset and the publicly available Microsoft Research Cambridge-12 Kinect (MSRC-12) dataset. On the first dataset we achieve an excellent accuracy rate of 96.72%. Furthermore, we show that Random Forests perform better than Support Vector Machines. On the second dataset we achieve an accuracy rate of 98.37%, which is on average 3.57% better then existing methods.

The motion sensor technology and the network of cameras have been explored separately as attractive solutions for building in-home monitoring systems for elderly care. Each of these technologies offers advantages over others while suffering from certain limitations. Motion detectors usually offer a privacy preserving solution, but do not yield granular information about the user's activities. Cameras, on the other hand, offer access to details of activities of daily life, but are regarded with caution in terms of coping with user privacy concerns. In this chapter, we provide an informative and a highly updated review of sensor fusion approaches. Then, we introduce an in-home monitoring system for elderly care, which is based on information collected from a network of PIR motion detection sensors and low-resolution cameras with 30 × 30 pixel arrays. The data fusion method we used for that system is based on different activity features. From the PIR sensors, the level of the daily occupation and the active level of the occupant are extracted. From the low-resolution cameras, other features such as activeness, sleep duration, visitors, and TV activity are extracted. Generally speaking, the proposed monitoring system fuses the heterogeneous sensor information to identify the occupant's lifestyle pattern. The system employs K-means clustering algorithm to group the observed days into different lifestyle patterns such as restful isolated days, restful social days, busy isolated days, and busy social days. To evaluate our system, experiments were conducted on six months of real-data. The results show promising performance to identify a certain lifestyle pattern or changes which occur over time.

This paper describes the development and testing of an Ambient Assisted Living (AAL) solution that combines state-of-the- art sensor technology with a social network application to empower elders to stay active, autonomous and socially connected and consequently support and unburden family caregivers. From a very early development phase both social scientists and engineers worked together to ensure a holistic approach to the development of the technology. To get a better insight into the needs, wishes and requirements of potential user groups, a qualitative approach was used to collect data among elderly and their caregivers from three European countries, namely Belgium, France and the United Kingdom. For the development of the technology, very low-resolution (900-pixel) visual sensors and Passive InfraRed (PIR) sensors were used and installed in a couple of homes of elderly to enable the monitoring and the analysis of the in-home activities of the elderly. In our application, automated analysis of both in-home activities and social network activities are fused to recommend new social connections and personalized activities to the elderly in order to stimulate social and physical activity. Additionally, algorithms for behavioral change analysis on the sensor data follow the health status of the elderly, such as the detection of disturbed sleep and eat patterns. From the qualitative studies interesting findings arose such as elderly mentioning safety and mobility benefits and the increased peace of mind while privacy issues, the lack of human touch and an unfelt need for support formed potential barriers towards adoption of AAL technologies.

The detection of foreground regions in video streams is an essential part of many computer vision algorithms. Considerable contributions were made to this field over the past years. However, varying illumination circumstances and changing camera viewpoints provide major challenges for all available algorithms. In this paper, a robust foreground background segmentation algorithm is proposed. Both Local Ternary Pattern based edge descriptors and RGB color information are used to classify individual pixels. Furthermore, camera viewpoints are detected and compensated for. We will show that this algorithm is able to handle challenging conditions and achieves state-of-the-art results on the comprehensive ChangeDetection.NET 2014 dataset.

This paper presents a new method to track multiple people reliably using a network of calibrated smart cameras. The task of tracking multiple persons is very difficult due to non-rigid nature of the human body, occlusions and environmental changes. Our proposed method recursively updates the positions of all persons based on the observed foreground images from all smart cameras and the previously known location of each person. The performance of our proposed method is evaluated on indoor video sequences containing person– person/object–person occlusions and sudden illumination changes. The results show that our method performs well with Multiple Object Tracking Accuracy as high as 100% and Multiple Object Tracking Precision as high as 86%. Performance comparison to a state of the art tracking system shows that our method outperforms.

In this paper, we propose a novel extrinsic calibration method for camera networks by analyzing tracks of pedestrians. First of all, we extract the center lines of walking persons by detecting their heads and feet in the camera images. We propose an easy and accurate method to estimate the 3D positions of the head and feet w.r.t. a local camera coordinate system from these center lines. We also propose a RANSAC-based orthogonal Procrustes approach to compute relative extrinsic parameters connecting the coordinate systems of cameras in a pairwise fashion. Finally, we refine the extrinsic calibration matrices using a method that minimizes the reprojection error. While existing state-of-the-art calibration methods explore epipolar geometry and use image positions directly, the proposed method first computes 3D positions per camera and then fuses the data. This results in simpler computations and a more flexible and accurate calibration method. Another advantage of our method is that it can also handle the case of persons walking along straight lines, which cannot be handled by most of the existing state-of-the-art calibration methods since all head and feet positions are co-planar. This situation often happens in real life.

Loneliness is a common condition associated with aging and comes with extreme health consequences including decline in physical and mental health, increased mortality and poor living conditions. Detecting and assisting lonely persons is therefore important-especially in the home environment. The current studies analyse the Activities of Daily Living (ADL) usually with the focus on persons living alone, e.g., to detect health deterioration. However, this type of data analysis relies on the assumption of a single person being analysed, and the ADL data analysis becomes less reliable without assessing socialization in seniors for health state assessment and intervention. In this paper, we propose a network of cheap low-resolution visual sensors for the detection of visitors. The visitor analysis starts by visual feature extraction based on foreground/background detection and morphological operations to track the motion patterns in each visual sensor. Then, we utilize the features of the visual sensors to build a Hidden Markov Model (HMM) for the actual detection. Finally, a rule-based classifier is used to compute the number and the duration of visits. We evaluate our framework on a real-life dataset of ten months. The results show a promising visit detection performance when compared to ground truth.

Unlike tracking rigid targets, the task of tracking multiple people is very challenging because the appearance and the shape of a person varies depending on the target's location and orientation. This paper presents a new approach to track multiple people with high accuracy using a calibrated monocular camera. Our approach recursively updates the positions of all persons based on the observed foreground image and previously known location of each person. This is done by maximizing the likelihood of observing the foreground image given the positions of all persons. Since the computational complexity of our approach is low, it is possible to run in real time on smart cameras. When a network of multiple smart cameras overseeing the scene is available, local position estimates from smart cameras can be fused to produced more accurate joint position estimates. The performance evaluation of our approach on very challenging video sequences from public datasets shows that our tracker achieves high accuracy. When comparing to other state-of-the-art tracking systems, our method outperforms in terms of Multiple Object Tracking Accuracy (MOTA).

With the rapid advances in mobile computing, handheld Augmented Reality draws increasing attention. Pose tracking of handheld devices is of fundamental importance to register virtual information with the real world and is still a crucial challenge. In this paper, we present a low-cost, accurate and robust approach combining fiducial tracking and inertial sensors for handheld pose tracking. Two LEDs are used as fiducial markers to indicate the position of the handheld device. They are detected by an adaptive thresholding method which is robust to illumination changes, and then tracked by a Kalman filter. By combining inclination information provided by the on-device accelerometer, 6 degree-of-freedom (DoF) pose is estimated. Handheld devices are freed from computer vision processing, leaving most computing power available for applications. When one LED is occluded, the system is still able to recover the 6-DoF pose. Performance evaluation of the proposed tracking approach is carried out by comparing with the ground truth data generated by the state-of-the-art commercial motion tracking system OptiTrack. Experimental results show that the proposed system has achieved an accuracy of 1.77 cm in position estimation and 4.15 degrees in orientation estimation.

Foreground background segmentation algorithms attempt to separate interesting or changing regions from the background in video sequences. Foreground detection is obviously more difficult when the camera viewpoint changes dynamically, such as when the camera undergoes a panning or tilting motion. In this paper, we propose an edge based foreground background estimation method, which can automatically detect and compensate for camera viewpoint changes. We will show that this method significantly outperforms state-of-the-art algorithms for the panning sequences in the ChangeDetection.NET 2014 dataset, while still performing well in the other categories.

© Springer International Publishing Switzerland 2015. Object recognition in video is in most cases solved by extracting keyframes from the video and then applying still image recognition methods on these keyframes only. This procedure largely ignores the temporal dimension. Nevertheless, the way an object moves may hold valuable information on its class. Therefore, in this work, we analyze the effectiveness of different motion descriptors, originally developed for action recognition, in the context of action-invariant object recognition. We conclude that a higher classification accuracy can be obtained when motion descriptors (specifically, HOG and MBH around trajectories) are used in combination with standard static descriptors extracted from keyframes. Since currently no suitable dataset for this problem exists, we introduce two new datasets and make them publicly available. De Geest R., Deboeverie F., Philips W., Tuytelaars T., ''Spatio-temporal object recognition'', Advanced concepts for intelligent vision systems - Acivs 2015, 12 pp., October 26-29, 2015, Catania, Italy. ispartof: pages:681-692 ispartof: ADVANCED CONCEPTS FOR INTELLIGENT VISION SYSTEMS, ACIVS 2015 vol:9386 pages:681-692 ispartof: Advanced concepts for intelligent vision systems - Acivs 2015 location:Catania, Italy date:26 Oct - 29 Oct 2015 status: published

Nearly half of the senior citizens report difficulty initiating and maintaining sleep. Frequent visits to the bathroom in the middle of the night is considered as one of the major reasons for sleep disorder. This leads to serious diseases such as depression and diabetes. In this paper, we propose to use a network of cheap low-resolution visual sensors 30 $$\times $$× 30 pixels for long-term activity analysis of a senior citizen in a service flat. The main focus of our research is on elderly behaviour analysis to detect health deterioration. Specifically, this paper treats the analysis of sleep patterns. Firstly, motion patterns are detected. Then, a rule-based approach on the motion patterns is proposed to determine the wake up time and sleep time. The nightly bathroom visit is identified using a classification-based model. In our evaluation, we performed experiments on 10i¾?months of real-life data. The ground truth is collected from the diaries in which the senior citizen wrote down his sleep time and wake up time. The results show accurate extraction of the sleep durations with an overall Mean Absolute Error MAE of 22.91i¾?min and Spearman correlation coefficient of 0.69. Finally, the nightly bathroom visits analysis indicate sleep disorder in several nights.

Copyright © 2015 SCITEPRESS - Science and Technology Publications All rights reserved. Analysing human sports activity in computer vision requires reliable segmentation of the human body into meaningful parts, such as arms, torso and legs. Therefore, we present a novel strategy for human body segmentation. Firstly, greyscale images of human bodies are divided into smooth intensity patches with an adaptive region growing algorithm based on low-degree polynomial fitting. Then, the key idea in this paper is that human body parts are approximated by nearly cylindrical surfaces, of which the axes of minimum curvature accurately reconstruct the human body skeleton. Next, human body segmentation is qualitatively evaluated with a line segment distance between reconstructed human body skeletons and ground truth skeletons. When compared with human body parts segmentations based on mean shift, normalized cuts and watersheds, the proposed method achieves more accurate segmentations and better reconstructions of human body skeletons. Deboeverie F., De Geest R., Tuytelaars T., Veelaert P., Philips W., ''Curvature-based human body parts segmentation in physiotherapy'', Proceedings 10th international conference on computer vision theory and applications - VISAPP 2015, pp. 630-637, March 11-14, 2015, Berlin, Germany. ispartof: pages:630-637 ispartof: Proceedings VISAPP 2015 vol:1 pages:630-637 ispartof: International conference on computer vision theory and applications - VISAPP 2015 location:Berlin, Germany date:11 Mar - 14 Mar 2015 status: published