An Algorithm for Vehicle License Plate Tilt Correction Based on Line Fitting Method

In the automatic recognition process of vehicle license plate (VLP), tilt correction is a very crucial link. According to the feature of K-Means Clustering (KMC), the image character coordinates are divided into two clusters (classes) to fit a straight line and then the line slope k is obtained. The rotation angle a is calculated using k and the whole image is rotated by α In the vertical tilt correction process, two correction methods, such as line fitting method based on KMC (LFMKMC) and line fitting method based on the least square (LFMLS), are proposed to compute the vertical tilt angle θ. After θ is obtained, Shear Transform (ST) is done to the rotated image and the final correction image is created. The experimental results show that, this paper algorithm (TPA), Compared with Hough transform, shortens the processing time and is more effective. Under the same condition, the processing time of TPA is 36 times faster than that of Hough Transform. It also provides a new effective way for VLP image tilt correction.