复杂背景中多车牌粗定位算法研究

提出了基于车牌分形维数特征进行复杂背景中车牌粗定位的方法。讨论了图像剪裁、灰度图转化以及图像增强时灰度转移函数的构造过程；给出了车牌图像分形维数的计算方法及车牌区域的确定。同时指出多车牌图像车牌区域的分形维数基本在2.65~2.80之间，其值高于车牌图像整体的分形维数，但是低于单车牌图像车牌区域的分形维数。该方法计算简单，不依赖车牌的颜色、形状、尺寸，具有极好的鲁棒性。通过对大量随机的实验图像进行计算表明：漏检率和误检率均为0，检出多于一个候选区域的为50%，正确检测率为100%。

Research on multi-vehicle license plate location under complex background

This paper presents a novel method of the vehicle license plate location based on characteristics of fractal dimension under complex background.Here the image clipping，the transformation of grayscale image，and the constructing procedure of grayscale transfer function with the image enhancement are researched.Then the calculation technique of fractal dimension of plate image is given meanwhile the license region is determined.Moreover，the fractal dimension of multi-vehicle license plate image is mainly from 2.65 to 2.80 which is larger than that of whole vehicle license plate image，but smaller than individual vehicle license plate image.Independent of color，shape and size of vehicle license plate，this algorithm not only is of convenience but also has good robustness.The performance of the proposed algorithm has been tested on a large number of experimental data from random and real images.Based on the experimental results，this algorithm shows both the missing rate and false detection rate are all zero.The possibility that the candidate region is more than one is 50%，at the same time the probability of correction for inspect is up to 100%.