Study On Vehicle Detection Technology Based On Line-CCD Camera

There are many video detection technologies, but the most is based on area CCD camera to obtain image. However, based on the area CCD, video detection technology is hard to solve some problems. (1) With the interested object of the video image, there are a lot of background objects, and in the image processing it is difficult to extract the interested ones from the complex background; (2) The field(frame)speed, and the pixel grad level are low, so that it is difficult to achieve the higher detection accuracy; (3) Data of video streaming is large amount (normally 10Mb/s), in order to ensure real-time, it is hard to use the complex image processing and analysis algorithms, so it is restrained to improve the detection performance. Compared with traditional video detection method of area CCD camera, it has the following advantages based on line-CCD camera: (1) It only contains the moving ones in the liner image sequences, because there is on background objects, it avoids the interference of the complex background; (2) It could be higher of spatial resolution, pixel-level and sampling rate of line-CCD camera;(3) Array of image data is quite small, so it is easy to achieve the complex image processing and analysis algorithms;This paper studies several key problems based on line-CCD imaging system, specifically, the major work may be summarized as followed:(1) Based on the study of conventional vehicles detection algorithm, it proposes a vehicle detection algorithm which is more applied on the line-CCD image. The algorithm is mainly based on the image texture, firstly using the wavelet transform extract the texture features from the road, which are as the basis for the binary versions, secondly it establishes some counters for each logical lane, and finally it segments the vehicle in accordance with the result. It is conformed that the vehicle detection algorithm is correct and validate by the experiment on-site, the algorithm can effectively eliminate the inference of the lights and shadows.(2) On the basis of vehicle segmentation, it achieves the object vehicle identification algorithm by projection method, the algorithm could realize vehicle recognition accurately.(3) Introducing the details of imaging software of SQL, it mainly implements video recording and playback, image segmentation, matching vehicles, license plate recognition, vehicle database management, data analysis vehicle, fault-tolerant capabilities.