Research of Abandoned Objects and Their Owner Detection Based on Omni-Directional Vision
|School||Zhejiang University of Technology|
|Course||Applied Computer Technology|
|Keywords||abandoned objects detection owner of abandoned objects detection omni- directional vision sensor abnormal behavior detection image moment|
The world anti-terrorism situation is grim. To ensure social security, especially in airports, stadiums, waiting room, exhibition hall and other public places, the detection of suspicious objects has become one of the most important contents of essential security systems.With the development of video surveillance technology, the underlying processing of image sequence such as image segmentation and background modeling has form a relatively complete theoretical system. On this basis, some domestic and foreign scholars have done a lot of research of abandoned objects detection and proposed various detection algorithms. However, there are some shortcomings in the existing video surveillance system: Can not display a wide range of scenarios without redundant information; Inadequate for determine rule of abandoned objects , thus prone to false detection; Once the abandoned objects have been detected, the video surveillance system can not promptly and effectively detect the owner of abandoned objects. For the above shortcomings, this paper propose, a detection system of abandoned objects and their owner, the main work and achievements were summarized as follows:1. This paper designed an experimental device by means of multi-vision sensor fusion of Omni-directional vision sensor (ODVS) and high speed dome camera. ODVS is use to detect abandoned objects and their owner, high speed dome camera is used to capture clear image of them.2. This paper proposed a method: Gaussian mixture modeling with two different update rate was used to detect the temporary static targets, image moment invariants was used to determine whether the targets are temporary static objects, finally, according to residence time and whether there are subjects near by, the relative objects were identified as abandoned objects.3. This paper proposed a method based on the flash back video to detect the relative owner of objects and confirm their current location by their trajectory.This paper propose, a detection system of abandoned objects and their subjects based on JAVA, describes and analyzes the realization of each module, series of relevant experiments were done in simulative environment. Experimental results show that the method used in this paper can effectively detect the abandoned objects and their subjects, with good accuracy and robustness.