Research on Abnormal Event Detection in Video Surveillance Based on Displacement of Feature Point and GLCM Texture Features

Abstract

The current anomaly detection algorithm uses more than a single feature to detect that causes trouble about the scene feature description and has the average accuracy of about 92%. This paper presented a new video anomaly detection algorithm using displacement feature and texture feature. The algorithm used pyramid LK optical flow method combined with Harris corner to extract the average displacement feature of moving target. GLCM algorithm and Fast Fourier Transform were introduced to get the texture feature of video images. Finally through SVM algorithms for two kinds of features classification training, the highest accuracy could reach 97.65%. Experiments show that the new algorithm has a higher accuracy and more practical.