基于形态学和梯度重构的车牌图像分割方法

Method for license plate segmentation based on morphological and gradient reconstruct

提出一种多尺度形态学检测边缘和梯度重构的快速车牌图像分割算法。该方法是基于多尺度的概念,进行梯度重构时采用不同尺寸的结构元素。运用开闭运算进行梯度重构,对重构后的各梯度图像进行阈值优化,对重构后梯度图像进行修改;利用分水岭算法检测目标,通过车牌的几何特性(宽高比、矩形面积等)从这些目标候选区域中剔除虚假牌照,找到真正的牌照。实验结果表明,该方法能取得较好的分割效果。

This paper proposed a method for fast license plate localization based on multi-scale morphological gradient optimization and gradient reconstruct.Considering the idea of muilt-scale,this method employed different sizes of structure elements to reconstruct morphological gradient image.Firstly,employed open and close operations to reconstruct gradient image.Secondly,extracted the minimal marker regional from each gradient image by optimizing thresholds and modified the reconstructed gradient image.Thirdly,performed the watershed transformation of the marker modified gradient image to achieve the regional segmentation of the image.Finally,marked the plate candidate regions and used the geometric properties（aspect ratio,rectangular area,etc.） to remove a false license to get a real license.Experimental results show that this method can achieve effectively segmentation.