XYZ空间与小波能量特征融合的车牌定位

针对复杂背景情况下的车牌定位问题,给出了一种融合了小波高频能量的方法。首先利用CIE-xy色品图进行颜色分割,找到符合车牌底色的候选区域。然后对候选区域进行数学形态学滤波,消除不必要的干扰。接着利用车牌特有的结构特征剔除明显不符合车牌特征的候选区域,提取符合条件的候选区域进行小波变换,由于车牌区域有着丰富的垂直方向纹理信息,因此比较候选区域的垂直高频能量,能量最高者即为初步选定的车牌区域。最后利用区域选择时的垂直方向小波变换系数对选择区域进行重构,并验证选择结果的正确性,如果为非车牌,则进行二次定位。该方法有效的解决了车身颜色与车牌底色相近时定位困难的问题。对各种条件下拍摄的225幅含有车牌的图像应用该算法,定位准确率达到98.2%。

Vehicle license plate location based on feature fusion of XYZ color spaces and wavelet energy

A new method to tackle the problem of complex backgrounds in license plate location,which is based on wavelet high frequency energy,is described.Color segmentation was first implemented by using an xy chromaticity diagram to identify the candidate regions.Secondly mathematical morphology was used to remove some of the interference regions.Thirdly,those regions which were clearly inconsistent with the characteristics of license plates were deleted and acceptable regions were subjected to a wavelet transform....