|
With the economy developing rapidly, more and more people focus on the traffic problem, safe, clear transports was expected by people. The Intelligent Transportation System (ITS) is a new topic which appears in this years, moving vehicle detection which based on video is a fundamental problem of video processing, which is also the basis of car tracking and car identification. In this thesis, we do some research about the key techniques of ITS and moving object extracting based on video. The main contents are as following:The images between daytime and night have large difference, so we should design different algorithm. We use the longitude and latitude to calculate the sunrise and sunset time of some place, then we can judge the day and night. With the judgement of day and night, we can use different algorithm.The state of signal lights which guide the cars are very important, which also can give us the signal to trigger the recording system. How to get the state of the signal lights is important, it means whether our system can work. In the thesis, we use the histogram, the gray variance and the range of gray value to judge the state of signal lights.In the daytime, there is sufficient sunshine, so we can use the difference of background to extract the objects. In this thesis, we use a new method to get the position of cars, which want to get the position of car plate, then we can get the tail of cars. In the algorithm of daytime, we also try to eliminate the shadow and person which affect our detection and get satisfy results.In the night, the light is not stable, so we can not get stable background. In this situation, we use frames subtraction and the edges of cars to extract cars. Another character is the tail lights of cars, we get the position of tail through the car lights, and do some check of the lights to ensure the correctness of results.
|