The Zebra Line Automatic Recognition And Early Warning

With the development of social economy, city traffic problem has been more and more concerned by people. Road intersections converge flow of cars and pedestrians, becoming the routes hub and core of the city road network where the traffic accident most likely happened.According to statistics, the United States had more than280traffic accident happened at the intersection or nearby in2000, which accounted for44%of all traffic accidents; The locations where death accident happened in Japan in1990shows that the number of accidents occurred in the intersection was near42.2%of the total number of accidents; Traffic accident statistics of China’s Ministry of Public Security in2003shows that traffic accidents happened in the intersection in city and country is about17%of the total number of traffic accidents. Thus, road intersection is very important to the traffic safety.Zebra crossing is an important safety sign in the road intersection. With the invention of the automobile, the city traffic becomes more crowded. For people crossing the street arbitrarily, traffic accidents occur frequently. Now, the zebra crossing recognition based on image processing has drawn the attention of researchers. Some foreign scholars have been studying on it since the beginning of twenty-first Century. However the recognition rate has not yet reached practical level. This is a key issue that restricts the traffic safety and intelligent traffic system.On the basis of previous research this paper proposed a zebra line recognition method based on improved Inverse Perspective Mapping (IPM). Four parts are included: the first part introduces and analyzes the traditional zebra crossing recognition algorithms; the second part, image acquisition and preprocessing. This part mainly introduced the methods of preprocessing and their effect on the zebra crossing recognition; the third part, IPM transform. This paper deduced the improved IPM transform applying the pinhole model of camera. Two classic IPM formulas were compared with; the fourth part, the zebra crossing recognition framework. This part describes the entire algorithm from the details including design criteria and its implementation. Finally, real world pictures are used to verify the zebra crossing recognition algorithm. Analysis to false-alarm picture is made and future improvement is presented.