2018 年 84 巻 857 号 p. 16-00366
An intelligent snatching detection system (ISDES) for autonomous detection of criminal incident by using machine learning approach is presented in this paper. Also, ESSM (Extended Snatching Situation Model) is presented to give an effective definition of the whole snatching situation including victim's behaviors. Rule-based decision making and machine intelligence are primary components of the proposed system to act without being explicitly programmed for data dispersion. Some motion characteristics are determined from video streams, also using the ESSM and machine intelligence, the system automatically classifies the situation of the video streams into criminal or non-criminal scenes. After constructing the system, we use test sequences that are continuous video streams of human behavior consisting of several actions in succession. We consider six types of scenarios for the experiments including victime's behaviors as the snatching incidents performed with motorcycle. The experimental results show that the system can effectively detect criminal scenes with an accuracy of 95.83 ％.
