The Tunnel Traffic Incident Detection Algorithm Based On Video

Detecting the traffic incident through the video sequence is an advanced method of trafficevent detection in intelligent transportation field at home and abroad. In current researches,improving the accuracy of event detection and then to lower the rate of false positives is apopular spot. This paper analyzes and researches the related video detection technology andthe key method, the domestic and foreign research present situation and the significance of thevideo-based traffic incident detection technology serve as a starting point. Highway tunnelstructure is mostly closed and complex, to ensure the safety of traffic inside tunnel is of greatimportance. This paper studies the road traffic tunnel events, the main research content is asfollows:(1) Researching and comparing the current moving object detection algorithm, andanalyzing the characteristic and complexity of video scenes inside the tunnel, choosing targetdetection algorithm based on depth difference accumulation because of the complicatedhighway tunnel traffic event detection. Introducing depth difference matrix to record thechange of the image in the process of the background modeling, the background is so accurate,clean, and of strong anti-jamming capability that can get accurate test results.(2) Coming up with a kind of event detection method based on adjacent steady-statevideo image difference. Using digital video and image processing technique to analyze andcalculate the grayscale video sequences. Then get the value of video image grey value, texture,variance value characteristics. And extract the corresponding steady-state characteristics fromthe digital characteristics. Analyzing the steady state characteristics to determine whetherthere is a traffic incident by looking for video sequences of two adjacent steady-stateparameters before and after difference. A large number of experiments show that the methodis of high detection sensitivity, low false positives, certain practical value and broadapplication prospects.(3) Analysis and research of all the current kinds of moving objects tracking algorithmtells the characteristics and applicable scope of the proposed algorithm. For small movingobjects, the tracking which based on area feature matching is fast and precise. The movingvehicles meet gaussian linear conditions closely, so the kalman filter is the most appropriate.