Enhanced Preciseness of Suspicious Activity Detection

Abstract

The major goal of the effort is to detect suspicious activities in surveillance video.  The approach created entails a number of stages of suspicious frame recognition and verification, as well as suspicious activity-related analysis of human motions inside a set of discovered suspicious frames. In the work presented here, different types of features are extracted to detect the suspicious activity. The technique includes GLCM feature extraction, which includes features like energy, prominence, contrast, entropy, and homogeneity type of features, matching using Euclidian distance, and descriptor features acquired by using Harris corner features and cosine similarity index estimation. The successful suspicious activity identification rate is examined, demonstrating a better performance and time-saving technique when evaluating a sizable collection of surveillance video.