基于改进Sobel算子的车牌定位方法研究

License plate positioning method based on improved sobel operator

针对在城市交通监控系统中车牌照定位要求的特殊性,提出一种基于多信息结合的车牌识别定位算法。该算法以摄像机拍摄的视频为基础,提取出视频中车辆的近景图像。根据车牌照图像特征,采用改进彩色Sobel算法对其作边缘检测,在处理后的图像中选取若干个车牌候选区。采用BP算法对车牌信息进行训练,将车牌面积、长宽比和垂直投影三个特征值作为输入值,进而达到车牌定位的目的。大量实验结果表明,该算法具有极高的定位准确度,具有重要的推广价值。

For the special requirements of car license plate positioning in the urban traffic monitoring system, a license plate location algorithm based on information fusion is proposed. The algorithm is based on video cameras, which extracts the vehicle close-up image from the video. According to features of car license plate image, using Improved color Sobel algorithm for edge detection, some plate candidate areas are selected from the processed image. Adopting BP algorithm to train the license plate information, the size, aspect ratio and vertical projection values of the plate are as input characteristic values, thus achieving the purpose of the license plate positioning. The numerous experimental results show that, the algorithm has a very high positioning accuracy and has important promotional value.