基于颜色分割与Sobel算子相结合的车牌定位

License Plate Location Based on Color Segmentation and Sobel Operator

在分析Sobel算子和颜色分割定位算法的基础上,针对Sobel算子在复杂背景及垂直边缘交错场景下定位效果不佳,以及颜色分割定位在车牌与车身颜色接近时导致无法分割的问题,提出了基于HSV颜色空间的颜色分割与Sobel算子相结合的车牌定位方法。通过将RGB颜色空间转为HSV颜色空间先进行颜色分割定位,当光照不均或车牌与车身颜色接近导致颜色定位失败时,再利用Sobel算子进行二次定位,并将获取的候选车牌送到已经训练好的SVM分类器模型,从而实现对车牌的精确定位。实验结果表明,与单纯的颜色分割定位和Sobel算子定位相比,该方法能进一步提高车牌定位的精度,并且在光照不均、车牌与车身颜色相近以及复杂场景下,都具有很强的适应性和鲁棒性。

Based on analysis of Sobel operator and color segmentation location,aiming at the problem that Sobel operator’s performance is poor under the complex background and vertical edge staggered scenario,and color segmentation can’t split the license plate with the body when their color is more close,we propose a license plate location method by combining color segmentation based on HSV with Sobel operator.Through transforming the RGB to HSV for color segmentation location,when color location is failed due to uneven illumination or similar color of license plate and car body,then the Sobel operator is used for secondary positioning,and the candidate license plate obtained is put into trained SVM classifier model so as to realize the accurate location of the license.Experiment shows that compared with the traditional color segmentation and Sobel operator,the proposed method can further improve the accuracy of license plate location and possess strong adaptability and robustness due to uneven illumination,similar color of license plate and car body and complex scene.