Crowd behavior understanding through SIOF feature analysis

Realizing the automated and online detection of crowd anomalies from surveillance CCTVs is a research-intensive and application-demanding task. This research proposes a novel technique for detecting crowd abnormalities through analyzing the spatial and temporal features of the input video signals. This integrated solution defines an image descriptor that reflects the global motion information over time. A non-linear SVM has then been adopted to classify dominant or large-scale crowd abnormal behaviors. The work reported has focused on: 1) online (or near real-time) detection of moving objects through a background subtraction model, namely ViBe; and to identify the saliency information as a spatial feature in addition to the optical flow of the motion foreground as the temporal feature; 2) to combine the extracted spatial and temporal features into a novel SIOF descriptor that encapsulates the global movement characteristic of a crowd; 3) the optimization of a nonlinear support vector machine (SVM) as classifier to detect suspicious crowd behaviors. The test and evaluation of the devised models and techniques have selected the BEHAVE database as the primary experimental data sets. Results against benchmarking models and systems have shown promising advancements in terms of the accuracy and efficiency for detecting crowd anomalies.