[Show abstract][Hide abstract] ABSTRACT: We propose an online anomal movement detection method using incremental unsupervised learning. As the feature for discrimination, we extract the principal component of the spatio-temporal feature by incremental PCA. We then detect anomal movements by an incremental 1-class SVM. In order to use principal component as the feature for discrimination while supporting incrementation of the subspace, we modify the SVM kernel function to take account of the difference in distance scale between the principal component feature vectors and that of the feature vectors after the subspace is incremented. This allows us to efficiently conduct the relearning process even though the dimension of the original input spatio-temporal feature is high. Experiments show that anomal scenes can be detected without the cost of preparing a lot of labeled data for preliminary learning.
[Show abstract][Hide abstract] ABSTRACT: In this paper, we present a new approach based on Markov Chain Monte Carlo(MCMC) for the stable monocular tracking of variable interacting targets in 3D space. The crucial problem with monocular tracking multiple targets is that mutual occlusions on the 2D image cause target conflict (change ID, merge targets<sub>hellip</sub>). We focus on the fact that multiple targets cannot occupy the same position in 3D space and propose to track multiple interacting targets using relative position of targets in 3D space. Experiments show that our system can stably track multiple humans that are interacting with each other.
[Show abstract][Hide abstract] ABSTRACT: We have proposed a method to detect and quantitatively extract anomalies from surveillance videos. Using our method, anomalies are detected as patterns based on spatio-temporal features that are outliers in new feature space. Conventional anomaly detection methods use features such as tracks or local spatio-temporal features, both of which provide insufficient timing information. Using our method, the principal components of spatio-temporal features of change are extracted from the frames of video sequences of several seconds duration. This enables anomalies based on movement irregularity, both position and speed, to be determined and thus permits the automatic detection of anomal events in sequences of constant length without regard to their start and end. We used a 1-class SVM, which is a non-supervised outlier detection method. The output from the SVM indicates the distance between the outlier and the concentrated base pattern. We demonstrated that the anomalies extracted using our method subjectively matched perceived irregularities in the pattern of movements. Our method is useful in surveillance services because the captured images can be shown in the order of anomality, which significantly reduces the time needed.
No preview · Article · Jul 2008 · IEICE Transactions on Information and Systems
[Show abstract][Hide abstract] ABSTRACT: We have developed a method that can discriminate anomalous image sequences for more efficiently utiliz- ing security videos. To match the wide popularity of se- curity cameras, the method is independent of the cam- era setting environment and video contents. We use the spatio-temporal feature obtained by extracting the areas of change from the video. To create the input for the discrimination process, we reduce the dimen- sionality of the data by PCA. Discrimination is based on a 1-class SVM, which is a non-supervised learning method, and its output is the degree of anomaly of the sequence. In experiments we apply the method to videos obtained by a network camera; the results show the feasibility of indexing anomalous sequences from security videos.
[Show abstract][Hide abstract] ABSTRACT: In this paper, we present a new approach for the stable tracking of variable interacting targets under severe occlusion in 3D space. We formulate the state of multiple targets as a union state space of each target, and recursively estimate the multi-body configuration and the position of each target in 3D space by using the framework of Trans-dimensional Markov Chain Monte Carlo(MCMC). The 3D environmental model, which replicates the real-world 3D structure, is used for handling occlusions created by fixed objects in the environment, and reliably estimating the number of targets in the monitoring area. Experiments show that our system can stably track multiple humans that are interacting with each other and entering and leaving the monitored area.
[Show abstract][Hide abstract] ABSTRACT: This work presents a new approach based on particle filtering to directly estimate the 3D positions of humans. Our system can predict occlusions due to other movements because we track humans in a 3D space, not on a 2D image plane. In addition, we introduce a 3D environmental model as the background model for tracking. This makes it easier to handle occlusions due to fixed objects in the environment. The 3D environmental model is automatically constructed by our original method from video sequences. Experiments show that our system is stable under occlusions due to the movements of both other subjects and fixed objects