Abnormal Behavior Detection Based On Optical Flow Features
Mesyella, Timotius Ivan Casey, Edward Susanto, Irene Anindaputri Iswanto · 2020
Crime is inevitable and unpredictable. It can happen everywhere at any given time. Fortunately with the advancement of technology, surveillance devices become more commonly installed in public places. CCTVs are strategically placed in various cities. These CCTVs are usually connected to the city control center. Monitoring all surveillance devices manually is impractical and may not be 100% accurate. There could be some crime activities and crowd incidents that go unnoticed. Thus, we are proposing a method to automatically and continuously detect abnormal, potentially against-the-law behaviours based on visual cues. We will use optical flow algorithms as the feature extractor. Then we will experiment on 3 different classifiers to find the most accurate and suitable classifier for this purpose.