自然场景下的车牌分割方法

摘要

The license plates extracted in natural scenes are easily to be incorrectly segmented due to the effects of complicated background such as inhomogeneous illumination,the tilt of angle and the motion blur.Aiming at the situation mentioned above, the plates were revised and binarized first,a new method combining the template matching and the dichotomy was proposed to segment the plates.The plate firstly was segmented using this method and then was adjusted with template to find the optimal segment point.The experimental result shows that the license plates can be segmented correctly using this method,the conglutination and breakage of characters can be overcome in different illumination situations.The robustness is excellent in natural scenes.%自然场景下的车牌面临着光照不均匀、角度倾斜、运动模糊等影响，易造成车牌分割错误。针对以上情况，对车牌进行旋转校正及二值化处理，提出一种基于模板匹配与二分法结合的方法对车牌进行分割。通过二分法进行分割，通过模板进行校准寻找最优分割点。实验结果表明，无论光照条件良好还是光线较弱或不均匀，该方法都能对车牌进行准确分割，有效克服字符粘连及断裂的情况，在自然场景下鲁棒性良好。