一种复杂背景下的多车牌图像分割与识别方法

License Plate Recognition from Complex Scenes with Multiple Cars

提出一种复杂背景下的多车牌图像分割和识别方法 .采用统计和特征匹配相结合的方法去除待识别图像中的背景 ,提取可能存在车辆的区域 ;分别对可能的车辆区域进行局部边缘检测 ,并使用车牌的先验知识确定车牌的位置和单个字符分割 ,包括车牌倾斜时的字符分割 ;使用 PCA和 BP神经网络相结合的方法精确识别车牌 .实验结果表明 ,该方法对复杂背景下多车牌的分割和识别是有效的 .

This paper addresses the recognition of multiple license plates from an image sequence acquired by a fixed video camera for video surveillance. The segmentation of cars from complex background is done using statistical method integrated with feature matching. Then license plate on each car is located and each single character is segmented. The single character recognition of inclined license plates by PCA and BPNN is also discussed. Experimental results show that using the method discussed, the segmentation and recognition of license plates in complex background can be done effectively.