Design And Implementation Of Embedded Vehicle License Plate Recognition System

With the increasing of our country’s vehicle, intelligent traffic management has become a kind of trend of development. License plate recognition technology plays a very important role in the intelligent traffic management, widely used in intelligent parking lot, high speed vehicle monitoring and electronic police. Aiming at low rate recognition and the high cost of the traditional license plate recognition system, the embedded license plate recognition system based on ARM is researched.This paper analyzes the background and current situation of research on license plate recognition, putting forward the overall design of the license plate recognition system. The design of an embedded linux intelligent license plate recognition system based on ARM is proposed. The hardware of the system adopts S3C2440 as the hardware platform, combining the software platform of linux operating system to realize image acquisition, image processing and other functions. In the license plate location, according to license plate based on rectangle feature and gray feature the pseudo plate region is to be removed, completing the exact location of the vehicle license plate. In the character segmentation algorithm using gray level jump off removes the frame, and according to the characteristics of the Chinese characters an improved vertical projection method is used to divide character segmentation. 21 points feature value extraction for the character recognized is proposed based on 13 point feature value extraction. The packet BP neural network character recognition is improved to recognize characters at last. In design of embedded linux system, it includes the establishment of cross compiler tool chain. According to the selection of the hardware platform U-Boot is made to transplanted and kernel is tailored, then file system is made. In the host system OpenCV is installed and used to write embedded license plate recognition system. Finally, the OpenCV library and the license plate recognition program cross compiled are transplanted on the ARM development board to run. The vehicle license plate recognition system is stable through the test.The results show that system adopts the improved positioning, the segmentation algorithm and the packet BP neural network recognition algorithm meeting the requirement of accurate recognition, realizing the automatic license plate recognition function.