In this paper, we propose an automated video surveillance system for kidnapping detection using featurebased characteristics. The localization of moving objects in a video stream and human behavior estimation are key techniques adopted by the proposed system. Some motion characteristics are determined from video streams, and using metrics such as a feature vector, the system automatically classifies the video streams into criminal and non-criminal scenes. The proposed system is called an intelligent security camera. We consider many types of scenarios for the training data set. After constructing the classifier, we use test sequences that are continuous video streams of human behavior consisting of several actions in succession. The experimental results show that the system can effectively detect criminal scenes, such as a kidnapping, by distinguishing human behavior.

Security camera Video surveillance Video processing Crime scene detection

Classification (of information) Crime scenes Motion characteristics Training data sets Video surveillance Intelligent security Video streaming Security systems Automated video surveillance Video processing Security cameras Video signal processing Cameras Social Sciences crime Behavioral research