An evaluation of object tracking performance using support vector machine and neural network for anomaly detection in crowded scenes

Abstract

The development of information technology and popularization of monitoring network and automatically detecting abnormal behavior in surveillance video is becoming more important for public security and smart city. The new approaches for moving object recognition are done based on context modeling. The existing Anomaly detection systems concentrate on spatial information yet neglect temporal information in part or in full, resulting in exposure to noise and background movement. Monitoring individuals in crowded scenes is a difficult task, owing to the variation in movement and appearance caused by the large number of people in the scene. An effective Kernel Support Vector Machine is developed for the anomaly detection utilizing spatio-temporal movement pattern models in overcrowded scenes to solve these problems. Initially, utilizing threshold value, the video is split into frames. The object movements are segmented using Extended Kalman Filters to improve the accuracy of the classification. The foreground and background object are identified using the texture features. Improved Learning Vector Quantization is used for efficient tracking of objects. Kernel Support Vector Machine is used to classify the anomalies. A machine learning and swarm intelligence-based approaches is developed for anomaly detection by extracting spatiotemporal features from video sequences. In this approach, the salient features are extracted from the frames to track and detect anomaly objects effectively. Initially, a two dimensional variance plane is constructed to encode local spatio-temporal variations around each pixel in a video frame. The Improved Particle Swarm Optimization algorithm is applied to isolate the most salient regions based on motion information in the two dimensional variance plane. A Grey Level Co-occurrence Matrix is then applied to the extracted salient pixels in the video. An Enhanced Artificial Neural Network based classifier is adopted to extract high-level features for abnormal event detection to improve the accu