Objectives: To detect interesting events in crowded environments by introducing Local Binary Patterns from Three Orthogonal Planes (LBP-TOP).The mission of automatically detecting frames with anomalous or interesting events from long duration video sequences has become the research interest in the last decade.
Methods/Statistical analysis: The existing system introduced a Swarm Intelligence based approach for Detecting Interesting Events in Crowded Environments. In this system both appearance and motion are measured to detect the anomalies. The Histograms of Oriented Gradients (HOG) is used for capture the appearance information and Histograms of Oriented Swarms (HOS) is used for capture the frame dynamics. Both are combined to form a new descriptor that effectively characterizes each scene. However it does not considered dynamic texture to achieve high accuracy. To solve this problem the proposed system introduced histogram of LBP-TOP to represent dynamic texture.
Findings: In a time window of each frame average triplets of HOG, HOS and LBP-TOP are consecutively computed. Then, these features are passed as an input to classifier. Here proximal support machine is used for classification. Proximal Support Vector Machine is based on Support Vector Machine, it is simpler and faster than traditional Support Vector Machines algorithm, which is especially suitable for large amounts of data or image classification and operations.
Improvements/Applications: The experimental results show that the proposed system achieves better performance compared with existing system.
Keywords
