Object detection and classification in surveillance system

Object Detection and Tracking in Surveillance System is inevitable in the present scenario, as it is not possible for a person to continuously monitor the video clips in real time. We propose an efficient and novel system for detecting moving objects in a surveillance video and predict whether it is a human or not. In order to account for faster object detection, we use an established Background Subtraction Algorithm known as Mixture of Gaussians. A set of simple and efficient features are extracted and provided to Support Vector Machine. The performance of the system is evaluated with different kernels of SVM and also for K Nearest Neighbor Classifier with its various distance metrics. The system is evaluated using statistical measurements, and the experiments resulted in average F measure of 86.925% and thus prove the efficiency of the novel system.