International Workshop on Background learning for detection and tracking from RGBD videos, ICIAP 2017, Sep 11, 2017
The detection of moving objects is the fundamental step in surveillance video analysis and meanin... more The detection of moving objects is the fundamental step in surveillance video analysis and meaningful toward object tracking and other higher level computer vision tasks. Nevertheless, many state-of-the-art methods are still limited in accurately detecting the moving objects because of complex background scenes such as illumination condition, color saturation, and shadows etc. Robust Principal Component Analysis (RPCA) models have shown potential for moving object detection, where input data matrix is decomposed into a low-rank matrix representing the background image and a sparse component identifying moving objects. Three deficiencies, however, still exist. First, one has to first restructure/transform the multi-way input video sequence into a data matrix. Such a preprocessing step usually leads to the information loss and would cause performance degradation. Second, RPCA models usually perform the matrix decomposition using batch processing. Therefore, it is very difficult for previous models to achieve real-time processing for disentangling moving objects from input scene. Third, previous models usually process color or intensity features for matrix decomposition. The lack of additional features also degrades the overall performance specially when color of background and foreground pixels is saturated. To address these problems, we evaluate the performance of Online Spatiotemporal Robust Principal Component Analysis (OS-RPCA) algorithm for moving objects detection using RGB-D videos. OS-RPCA is a graph regualrized algorithm which preserves the low-rank background spatiotemporal information in the form of dual spectral graphs. One graph is constructed among the columns of data matrix and second graph is constructed among the spatial locations of the data matrix. A novel objective function is designed which encodes the spatiotemporal graph regularized constraints. The objective function is solved using online optimization scheme.We evaluate OS-RPCA algorithm for moving object detection on new RGB-D dataset, and show competitive results as compared to others state-of-the-art methods.
Bookmarks Related papers MentionsView impact
Uploads
Papers
large number of studies on convolutional neural networks applied to background subtraction have been published, and a continual gain of performance has been achieved. In this context, we provide the first review of deep neural network concepts in background subtraction for novices and experts in order to analyze this success and to provide further directions. To do so, we first surveyed the background initialization and background subtraction methods based on deep neural networks concepts, and also deep learned features. We then discuss the adequacy of deep neural networks for the task of background subtraction. Finally, experimental results are presented for the CDnet 2014 dataset.
large number of studies on convolutional neural networks applied to background subtraction have been published, and a continual gain of performance has been achieved. In this context, we provide the first review of deep neural network concepts in background subtraction for novices and experts in order to analyze this success and to provide further directions. To do so, we first surveyed the background initialization and background subtraction methods based on deep neural networks concepts, and also deep learned features. We then discuss the adequacy of deep neural networks for the task of background subtraction. Finally, experimental results are presented for the CDnet 2014 dataset.