Tilt Plate Image Correction Algorithm Based On DSP

License plate recognition technology in many areas of modern society in an increasingly wide range of applications, to facilitate the development of social life and bring huge economic benefits. License plate recognition system, license plate tilt correction is an important step link plate positioning and license plate character segmentation, can make the impact of this step of the entire system is particularly large, fast, accurate calibration work to do tilt plate image, able to subsequent license plate character segmentation to provide effective protection, to enhance the accuracy of the license plate recognition has a significance.This paper studies the tilt correction wavelet transform and improve the structure of DSP-based hardware algorithms to tilt correction Fourier transform algorithm as a transition, were studied hough improved algorithm based on wavelet transform and wavelet transform algorithm based on improved radon. Approach of the algorithm is:first the collected license plate tilt wavelet transform image; then transformed image in its tilt correction. In order to verify the feasibility of the algorithm, the paper has done a lot of simulation and experiments, including the effects of different wavelet decomposition level of correction effect of research to improve the algorithm for different pictures of universality, and the effect of different wavelets on the correction effect;Experimental results show that the improved algorithm based on wavelet transform this relatively new algorithm to remove the license plate tilt, has certain advantages compared to the traditional method:First, make full use of the advantages of DSP image processing in real time, so that the algorithm used in the actual case has the speed, large capacity, small size, light weight; secondly, the algorithm of this paper can overcome the traditional method of the plate over the border in accordance with patience, overcome by noise pollution is difficult to tilt the plate image correction features. Finally, because of the change in the wavelet algorithm performed under the sampling process allows to significantly reduce the computation time, real-time stronger.