The Design Of Embedded Vehicle Plate Recognition System Based On DM642

ITS(Intelligent Transportation System)is the main development direction in the traffic management, and the LPR(License Plate Recognition )is one of the key technologies of ITS.License plate recognition system is consisted by three parts: License Plate location, character segmentation and character recognition. In order to improve the real-time and the identification accuracy of the license plate system, this paper proposed a new idea which designs an embedded vehicle plate recognition system based on the DM642. Combining the high performance DSP chip DM642 with the license plate recognition algorithm, the performance of the system can be improved greatly.In this paper, we have studied license plate recognition algorithm. In the aspect of license plate location, this paper proposes a new algorithm that regional density scanning method, which combining with morphology can locate the license plate more accurately.Regional density scanning method is to divide the image into several regions, using the unique character density features of license plate regional to delete some interference regions. Dividing the image into several regions, and then scanning, this can improve the processing speed and accuracy.The method of regional density scanning and the morphological processing has a higher accuracy than the method based on edge detection and the morphological processing.For the character segmentation, we improved the iterative binary algorithm and tilt plate correction method. In the iteration binary algorithm, the fixed weight is 0.5, which can't adapt the unstable light of the plate. So, in the paper we make the weight be adaptive to meet the changeable situations. And the effect is very obvious. Slant Correction of the tilt license plate, we use the simple geometric method that according to the level projection difference between the front and back of the plate to compute the tilt angle. In character recognition, we extract the closed-loop feature before template matching. According to the characteristics of the closed-loop, we make a preliminary classification. This method reduces the computation time and improves the recognition rate.Finally, this paper transplants the license plate recognition algorithm to DM642 platform with comparison of the difference between the vc++ platform and CCS development environment of the algorithm.In a word, embedded vehicle plate recognition system based on DM642 can realize a good performance of license plate recognition, and got some improvement both in recognition and recognition rate.