The vehicle license plate localization based on image processing, domestic and foreign scholars have done a lot of research, but still not enough perfect. The existing algorithms can handle only some specific situation and can’t completely adapt to various factors affecting the license plate location, which is the main problem. These factors include the changing environmental background and illumination conditions, license plate different size and different degrees of inclined in image. This paper made a great improvement in this respect.This paper uses gray image as the processing object, using the improved Sobel edge detection operator and the Otsu threshold binarization, as much as possible to retain the license plate edge feature, at the same time to filter out invalid interference, then make use of horizontal projection combined morphological dilation and erosion processing steps to form the connected area to roughly locate the license plate. The advantage is that the structure of choice can depend on the height determined by horizontal projection and automatically adapt to the change of license plate size.The coarse location need of license plate tilt correction. In view of the Hough trans-form computation time consuming and the shortcomings of the poor stability, this paper compute the angle of license plate by the method of least squares fitting, and two kinds of algorithm was compared. Experiment shows that this method is not only fast in calcu-lation speed, but also have strong anti-jamming and high accuracy. Finally license plate image are correctly rotated with bilinear interpolation and precise cut off further more. |