Video Processing Method And System On Intelligent Traffic Surveillance

In this modern society, as the developments of technologies the scale of roads continues to expand, and travelling of people becomes more and more frequent. In the same time, it brings up some negative effects. The data shows that each year the amount of traffic accidents in the world are increasing. In order to reduce the impact of these issues on people’s travelling, a wide variety of products in Intelligent Traffic Systems (ITS) came into being. In these products the video surveillance systems are the most valued. Compared to the traditional sensors monitoring methods, the video surveillance is simple and convenient and it’s the new direction of the ITS in the future. In this paper, we study the methods of traffic video processing and system construction.In this paper, we take a method based on the background subtraction to make image target detection. Compared to the optical flow and inter-difference method, background subtraction is more simple and accurate. We take the multi-frame averaging algorithm to subtract the image background because it has higher computational efficiency. In addition to using the conventional real-time background update algorithm, but also we presents a background update algorithm to detect the background mutation based on the number of fore targets. It achieved good results in practical applications.After the extraction of foreground, we give a target tracing algorithm based on inter-overlapping area by using the characteristics of traffic targets. It can accurately track the history positions of each target in the area.We take a research on the illegal surveillance of traffic videos. We gave the criteria and formulas of some illegal actions, like crossing the yellow line, reverse driving and emergency lane occupancy. By analyzing the picture target track data to determine whether it is illegal and which action it is doing. The judging criteria tested with high accuracy.Aiming at the non-straight road this article also does some corresponding algorithm optimization and improvements, like road modeling, the simulation of the boundary on the road, vehicle trajectory prediction optimization etc. These optimizations make the program more extensive range of applications.This paper presented a real-time traffic monitoring system building method. We constructed the traffic surveillance platform by using some hardware devices like HD Dome Camera, industrial control computer and 3G transmit modules. This platform can achieve functions like real-time data analysis and vehicle monitoring and so on. In the vehicle monitoring we gave a Vehicle Trajectory Prediction Snapping algorithm by using the vehicle history track. It makes the snapping pictures more accurate and clear. The algorithms and program developing in this article is both in the MFC platform of Visual C++ 6.0. After the actual application environment test of the program, the program achieved some satisfactory results.