The Methods Of Vehicle Information Video Extracting And Their Application

Vehicle detection is a basic question in Intelligent Transportation System (ITS). And the video detection technology, which is one of the important research fields, will push ITS ahead. It has been greatly used in traffic monitoring and controlling, automatically tolling and so on. However, there are lots of problems to be studied for the complexity of the video detection. The vehicle classification is one of the key techniques. The thesis worked on the underlying aspects from the following points.The thesis researched on a two-dimensional (2-D) thresholding by a genetic evolution mechanism. The 2-D thresholding can improve the segmentation quality. However, its computational speed is so slow that is difficult to be used greatly. Genetic Algorithm (GA) can accelerate the threshold selection, but the premature convergence must be considered. This thesis proposed a novel genetic evolution mechanism based on schema. By introducing this mechanism, the search efficiency would be improved in each iteration cycle.The thesis proposed a two levels vehicle classification method. Firstly, some moving vehicle characters would be extracted, such as the shape complexity, the ratio of top width and total width of vehicle, to implement the shape identification roughly. Secondly, the vehicle could be classified by wheel number and their distance furthermore. For the sake of the low detection speed, proposed a Hough transform by the continuous curve to detect wheel. A new method, which would reduce the invalid sampling accumulation, was presented to improve the random Hough transform method on circle detection.The thesis studied the moving vehicle detection by background difference. Considering the park characters, a self-adaptive background difference method was used to extract the movement target.The thesis developed a park management demo by video detecting technique. And this demo was developed by some of aforementioned research.