Research on Vehicle Detection Method Based on Video Image

摘要

Nowadays, there are quite a few methods that have their respective advantages and disadvantages to update the background, such as method of multiframes average, method of Gaussian distribution models etc.. Considering those disadvantages above, the background updating algorithm which be advanced in the research can reduce the effect by shadow effectively by doing the calculation of H chrominance component on HSI space, achieving a lasting update automatically by operating the frequency statistics after every sampling. Applying the method of combining edge detection of morphologic and difference of background together to the target detection, a method of foreground area extraction based on the edge information has been brought forward in this research. The method of morphologic edge detection algorithm has a good restrain on noises while doing the edge detection. Therefore, in vehicle detection, the background difference will be done after respective morphologic edge detection of the current video image and background image. Use the extraction template of background edge to extract the precise background edge of the current frame, then a foreground edge of the vehicle will be extracted, finally use the mathematical morphology to do the later treatment to the result of target dividing, remove the noise. The experiment has proved that the measure mentioned above has increased the accuracy and stability of vehicle detection effectively.