Research On Detection Algorithm Of Abandoned Object In Complex Environment Based On Vision

Fighting against terrorism and ensuring public safety have gradually become the focus around the world after the September 11 attacks. Now,the main means of terrorist attacks are used by lefting objects on stations, shopping malls and other sensitive public areas. As the development of technology in twenty-first century,people requests more for security facilities, and intelligent video surveillance system are widely used. Detection and recognition of abandoned objects are key parts of intelligent video surveillance systems. At present, the detecting and segmention algorithms based on vision is relatively rare in complex environment, the main reason is the forground objects blocks heavily to the background objects, and coming with much noises'interference, so it is difficult for this study to some extent.This paper mainly contains researches on detection algorithm of abandoned object in complex environment. The complex environment is that the foreground object have higher density, move arbitraryly, block each other and have large noise's disturbution with the change of weather and light. First, we presents an algorithm on detecting the abandoned objects based on Temporal Median Image Sequence, it can handle the serious occlusions and noise of scene in complex environment caused by the motion of foreground objects,At first, the algorithm get the TMIS by taking the median filtering on the surveillance video image sequence to segment on time dimension, then use the logic operation between the TMIS and the foreground object image in corresponding,we can find the candidate sequence of abandoned object. Our method is robust to occlusions and noise interference in complex environment.Second, this paper describes an algorithm of Regional Match that segment the abandoned object.It mainly solved the problem of judging and segmentation to the candidate sequence which has been detected in the first step. This algorithm use the open operation and blob analysis, and then store the candidate sequence of abandoned object and its information composed of one element into the array of regional structures,and then find the real object by compare the Regional Match on which elements and subsequent frames of the structure elements with their various criterion and threshold. It can reduce the missing rate and false positive rate of recognizing the abandoned objects. Furthermore, the algorithm has low time complexity.The result of experiment shows abandoned object in complex environment have high detection and recognition accuracy.Compared with other algorithms,the algorithm of this paper have improved the efficiency on detection and identification,and the complexity of detection and segmentation algorithm is more stable in time, it meets the real-time requirementsts.