Detection and Recognition of Abnormal Behaviour Patterns in Surveillance Videos using SVM Classifier
4 Pages Posted: 16 Sep 2019
Date Written: September 14, 2019
Abstract
In recent day’s video surveillance is important to identify the abnormal activities of human in order to maintain serene environment. This paper proposes an automatic detection of abnormal events in the recorded video. This work mainly focuses on detection of human (Student) and their abnormal activities in academic zone. The proposed system consists of three phases 1. Motion Segmentation 2. Feature Extraction 3. Action Classification. Background subtraction method primarily used to segment the moving object. Feature extraction is carried out by Hu moments. Classification of normal and abnormal activities is done by support vector machine (SVM) classifier algorithm. The proposed method evaluated with the recorded video dataset “SAIAZ” (Student Activities In Academic Zone) of our Institution and “Sarkar” tamil movie dataset. The SAIAZ and Sarkar” tamil movie dataset consists of activities like slapping, kicking, punching, hitting, and fighting. The performance metrics such as TPR, TNR, FPR, FNR are used to evaluate the proposed system. The accuracy of action detection rate in SAIAZ is 72% and “Sarkar” tamil movie dataset is 68%.
Suggested Citation: Suggested Citation