车牌定位与字符分割技术研究-电子与通信工程专业论文.docx

I I 摘 要 车辆牌照自动识别系统在桥梁路口自动收费、停车场入口管理等领域有着广泛 的应用。车牌的自动定位分割是车辆牌照识别系统的关键技术之一。由于车辆图像 采集于户外，图像背景复杂，噪声干扰严重，图像质量较低，因此牌照的自动定位 分割一直都不是很理想，使得它一直是该领域的研究热点。 本文通过对大量相关资料的搜集、整理、分析，总结了近年来国内外在车牌定 位分割领域的最新研究成果和最新进展，在此基础上，选择了一些图像预处理的方 法对图像进行预处理，得到适合后续定位的高质量图像；本文详细分析了车辆图像 的特征，尤其是牌照区域的特征，通过选择二值化图像中黑白像素点高频变化的连 通区域来定位车牌。考虑到车牌可能倾斜，文中引入了 Hough 变换进行倾斜角的检 测，校正倾斜车牌。本文对分割出的车牌图像进行灰度拉伸及直方图均衡化处理， 并对整个车牌进行了背景色与前景色调整，采用最大类间方差求阈值法处理车牌图 像。然后精定位车牌，扫描车牌区域每列白像素点，结合车牌字符本身的几何特征 进行字符分割，使最终的字符块能满足后续字符识别的要求。本文对实验结果进行 了统计和分析，并在文章的最后对后续的工作和算法前景进行了展望。 本文对 150 幅车辆图像进行一次粗定位，二次粗定位和字符分割。经实验证明， 所采用的算法能够有效的定位，准确地分割字符，系统的实时性及准确率都能达到 既定目标，因此具有较好的实际应用价值。 关键词：车牌识别，图像预处理，车牌定位，倾斜校正，字符分割 II II Abstract Automatic vehicles’ license plate detection and recognition system can be broadly used in bridge intersection’s automatic charge and park’s entrance management. The automatic detection of license plate is one of the key technologies in LPR system. Because a license image is taken outdoors, the image background is complicated, and there is serious disturbance in the image so that the image quality is bad. The location of plate is not good enough. Therefore, the study on the algorithm of license plate location is always the hotspot problem. In this thesis, I have collected large numbers of interrelated data, and then summarized the latest research achievements and development of license plate location and segmentation. Then I analyze some image disposal measures, and then choose some measures to deal with the images, so we can get good images to locate the license plate. I analyze the characteristic of license images, especially in license plate area, and then defined the most change in pixel value as the most important parameter. Considering the lean of vehicle license plate, I adopt Hough transform to detect the lean angle and rotate the images to made some compensation. Then I enhance and proportion the grayness value of the license plate, and then rectify the color of the license plate. Finally, I locate and segment the vehi