一种基于多特征提取的实用车牌识别方法

针对车牌识别系统的实际应用, 利用车牌区域的边缘梯度特征、几何形状特征、颜色特征、灰度纹理特征定位车牌, 然后校正车牌图像的颜色及倾斜度; 基于灰度投影法, 对普通及武警车牌均提出了有效字符分割方案, 通过自适应判别去除因字符断裂粘连、特殊字符等造成的干扰; 通过基于多特征值提取的神经网络方法初识别车牌; 最后将人眼的视觉特性用于模板匹配法, 解决易混淆字符及污损车牌的问题。通过大量实验证明, 该方法对车牌颜色、拍摄角度、光照条件等限制较少, 适用范围广、识别率高, 有较强的实用性。

Practical license plate recognition method based on multi-feature extraction

According to application of LPR system, this paper used the edge features, geometry features, color features and texture features to locate the LP, then realized LP's color and skew emendation. It proposed a segmentation method for both ordinary LP and armed police LP, removed the interferences caused by adhesion and some special characters based on adaptive judging mechanism. It realized initial recognition by using multi-features based on neural network and used human's visual mechanism for template matching, which could recognize some stained LPs and some characters in similar structures. Much experimental results show that this method is effective, which has wider application range and overcomes the sensitivity to image environment, illumination, color and shooting angle.