Research Of Vehicle Detection Technology Based On Embedded Vision System

Road traffic injuries take an enormous toll on individuals and communities as well as on national economies every year because of the increase in the number of registered vehicles and complicated traffic states. With the concept of active safety, especially Advanced Driver Assistance Systems(ADAS), traffic safety has a new development direction. Vehicle active safety has been research focus because of it is able to resolve traffic problem more efficiently.This paper presents an adaptive vehicle detection algorithm based on embedded vision after analyzed the research achievement of vehicle detection at home and abroad.Firstly, this paper designed embedded hardware platform. The OMAP3530 dual-core microprocessor, belongs to TI’s DaVinci technology, is adopted in the hardware platform. This paper also designed image collecting module and display module, and configured the system’s basic application. The hardware platform provides the excellent hardware basis for porting algorithmic program.Secondly, for the highway or high grade highways, this paper provides a vehicle detection algorithm which can not only detect vehicles during the daytime, but also work in different lighting and environmental conditions at night. Firstly, the driving environment is classified into four types by creating environmental classifier:high light illumination, normal illumination, low light illumination and nighttime. Then, presenting the method of edge region growing is aimed to divide environment into non-road and road area; and in the light of the different classifications of environmental conditions. Finally, the algorithm detecting preceding vehicle based on multi-feature fusion is put forward.Thirdly, this paper takes version selection, structure cut, function rewriting and other measures to port OpenCV’s API to DSP core according to the analysis of the characteristics of OpenCV and DSP. This paper has ported the proposed vehicle detection algorithm to embedded vision system, at the same time, keeps the accuracy and diversity of OpenCV function.Finally, applying the algorithm presented in this paper to the road test in strict accordance with strong light intensity, Normal illumination, low light intensity and night, and statics miss detection rate and false detecting rate. The testing result shows that the algorithm presented in this paper possesses strong environmental adaptability so that the preceding vehicles can be detected quickly and accurately...