Moving Objects Detection And Warning Technology For Smart Traffic Surveillance System

In recent years, the application of Intelligent Traffic Surveillance System (ITSS)has been paid more attention, it is the use of computer vision technology, analyzing ofvideo data to get meaningful information, to provide effective help for trafficmanagement. Traditional surveillance system requites watchers to gaze at the monitorto subjective judgments at any time. However, with the increasing number ofsurveillance equipment, the stress and burden of watchers have been heavier. ITSScan automatically analyze the sequence video images and extract moving objects inthe video. By analyzing the behavior of moving objects in the scene, ITSS can make atimely response to the abnormal phenomena, and real-time outputting warningmessage to watchers to deal with the situation in time, which can let the traffic returnto normal and reduce the loss of property related personnel.This paper mainly studies the following three aspects:Firstly, Moving Objects Detection (MOD), detecting moving objects in videoframe. In this paper, we propose a simple and effective approach to detect movingobjects in video images—Space Vector Difference (SVD). The method utilizes theSVD between current video frame and background model, which has two importantproperties: length and direction. Then, by observing the statistical characteristic of thelength and direction, the moving objects and noises are classified, and an adaptivethreshold is automatically calculated to achieve the automatic detection of movingobjects and removing of its noise. Finally, for the isolate noise points, they can beeliminated by the method of mathematical morphology, and the incomplete objectswere repaired. The experiment results demonstrated that the final detection resultswere good, which can provide safeguard for the next target tracking.Secondly, for objects tracking, extracting the feature information of the objectswith the detection results at first. Then the paper introduces the basic principle ofKalman filter and application in tracking. In the target tracking process, generally,target tracking just take the overlapping area between the bounding rectangles ofobjects as the basis of judging, algorithm is simple. But, for some complex situations, such as objects synthesis, not only using the overlapping area which cannot meet therequirement of tracking, but also using Kalman filter to predict the target location andarea next time. Then, we can determine exact location of the target and update thefeature data of the target with the template matching within the area.Finally, the early warning technology is explored in ITSS. We can get the speedand direction that are the key motion parameters of the target based on analyzing thetarget behavior. This paper discusses three abnormal phenomena in detail in trafficsurveillance video: traffic congestion, pedestrians crossing the road and trafficaccidents, and also presents the related detection algorithms. The algorithm hasrealized with ideal effect and can meet the requirements.