Title: Real-Time Detection of Suspicious Human Movement

Year of Publication: Nov - 2014
Page Numbers: 56-69
Authors: Sugla Vinaayagan Rajenderana, Ka Fei Thang
Conference Name: The International Conference on Electrical, Electronics, Computer Engineering and their Applications (EECEA2014)
- Malaysia

Abstract:


Crime is everywhere and it could be argued that we are in one of the most crime eras in human history. Crimes like theft, violence against people and property damage are some of the rising crimes in university. On the other hand, investigation is held to find the person who is responsible for the theft, only after the crime occurred, by using existing surveillance system. It acts as 'post-mortem' tools. Therefore, in this paper, an appropriate algorithm for autonomous suspicious human-movement detection from surveillance videos is proposed. First, proposed system extract human-movement information by detecting and tracking people in real-time using Gaussian Mixture Model (GMM). Morphological operations aid the detection of human-movement by eliminating noises. Features extracted from human-movement are then sent for post-processing in order to recognise whether the detected activity is suspicious or not. Examples of suspicious movements are loitering and hanging or looking around in the area of interest for longer time period. The framework used for recognising suspicious activities is called Grammar-based approach. This approach is effective in detecting suspicious activities with serial of events for a longer time period. By using recorded videos with people mimicking those suspicious activities, several experiments have been performed and results presented in this paper. The experimental results presented here demonstrate the outstanding performance and low computational complexity.