基于识别结果校正的车牌字符分割算法

在车牌识别系统中,字符分割发挥着承上启下的作用。由于外界环境如光照、尘埃的影响,造成预处理后的车牌图像产生字符断裂和粘连现象,造成分割错误,影响后续字符识别。通过研究多种字符分割算法,提出了一种结合垂直投影法、连通域分析以及通过模板匹配字符分类器的识别结果进行反馈校正的分割方法。分别对停车场车辆车牌、高速公路路口抓拍的车牌以及公路行驶车辆抓拍的车牌进行测试,其中停车场车牌可以实现98%以上的分割正确率,其余2种情况的正确率也在97%左右,表明该方法有实用性。

License plate character segmentation algorithm based on the calibration of identification results

Character segmentation plays a transition role in the license plate recognition system. The external environment such as light and dust may cause the character fracture and adhesion after the pretreatment of license plate image, making segmentation error and affecting the subsequent character recognition. Therefore, by studying a variety of character segmentation algorithm, a segmentation method is proposed by combining vertical projectionwith connected domain analysis and a feedback correction method by the recognition results of the template matching classifier. This method is respectively tested on the vehicle license plates which are captured on the parking lot, the highway intersectionand the road vehicles. 98% of segmentation correct rate can be acquired in the first case while around 97% for the other two, showing that the method is practical.