Design And Implementation Of Vehicle License Plate Recognition System Based On Cloud Platform

With the rapid development of Intelligent Transportation Business, the data of traffic monitor have increased explosively. The traditional monitor and recognition systems have been unable to meet the need of high-speed processing of massive high-definition images and video data. In order to solve the problem, this paper designs and implements the License Plate Recognition System which is based on Cloud Platform by combining Open Source Cloud Computing Platform, License Plate Recognition Technique and Video Processing Technology. Comparing with the traditional Single-System, the experimental results shows that this system has more powerful capability for data processing and higher computational efficiency. The main contents are as follows.(1) The Research, construction and performance optimization of Hadoop Cloud Platform. In order to realize distributed processing of huge data, in this paper, on the basis of in-depth study of Hadoop Cloud Platform, we focus on principle of operations and implementation methods of the MapReduce framework. The Hadoop Cloud Platform is set up on four computers. This paper deeply analyzes configuration parameter of platform and gives the optimal configuration plan. The result shows, after optimization of Hadoop platform, speed of both image processing and video processing is promoted 4 times and a quarter respectively.(2) The simulation and analysis of the License Plate Recognition Algorithm. This paper introduces the algorithm of three parts which are license location, character segmentation and character recognition. It designed a complete identification process and then has made a comparative analysis and simulation verification among these chosen algorithms during the process.(3) The design and implementation of distributed processing module. This thesis expands the data type of Hadoop and the design implements two functional modules: 1.The pattern recognition module based on MapReduce, including the license plate recognition and the function of image with shooting information; 2.The video recognition module based on MapReduce, including the automatic video segmentation, the frame extraction, the license plate recognition and the function of the time orientation. Among these, to get around the problem that the Singe works in a low efficiency, the method that a recognition of distributed image and the extraction of video frames of distributed images has been designed, which improves the computing power of system. In order to solve the problem that the manual searching has the low efficiency, it designs a method of time for orientation by videos which has implemented the rapid test and time orientation when the vehicles appear in the video.(4) The license plate recognition system based on Cloud Platform has been implemented as well as analyze its properties. The license plate recognition system has been implemented on the Hadoop platform. That has completed the interaction design based on the system of the Web. We made a test for each module of system and the result showed, comparing with Single System, under the recognition system based on Cloud Architecture, the recognition speed of image and video processing has improved more than three times and two times. Moreover, it enables to handle the large videos over than 2G more rapidly and steadily.