摘要： 基于灰度图像水平差分和垂直差分图像的相关特点，提出了一种能在复杂交通背景下实现车牌快速定位和分割的方法。该方法首先通过中值滤波对图像滤噪并进行直方图均衡化处理；然后运用一种水平差分和自适应阈值方法对图像进行二值化，接着在对所得二值图像在水平方向作形态学膨胀运算后，对其垂直、水平方向的扫描线长度先后进行滤波处理而粗分割出车牌的候选区域；最后通过车牌垂直差分图像在竖直方向扫描投影的相关特征对所得候选区域进行筛选，得到车牌在水平和垂直方向的精确位置。该方法对于图像背景复杂度和车牌在图像中的位置限制较少，而且实现简单，定位速度较快，是一种简单、高效的车牌定位方法，实验结果证明了该方法的有效性。

关键词: 车牌定位；水平差分；垂直差分；扫描投影

Abstract: Based on related characteristics of the horizontal and vertical difference results of gray image, a method to obtain location and segmentation of vehicle plate under complicated traffic background is produced. At first, image noise is filtered by mid-filter after process of histogram equalization, then candidates of vehicle plate is coarsely segmented by filtering scan line length orderly in the horizontal and vertical direction of the binary image after operation of horizontal dilation, the precise position of vehicle plate in the horizontal and vertical direction is selected out among the plate candidates at last according to the related characteristics of scanning projection of the difference image in vertical direction. With less restriction on background complexity of the image and the plate position, the presented method can be applied with simple and fast implementation, whose validity is proved by the experimental result.

Key words: Vehicle plate locating; Horizontal difference; Vertical difference; Scanning projection