Font Size: a A A

The Research Of Vehicle License Plate Recognition And Designed On GPU And CPU

Posted on:2012-09-03Degree:MasterType:Thesis
Country:ChinaCandidate:F WangFull Text:PDF
GTID:2178330335993349Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
With the raising of vehicles and the worsening of urban traffic conditions, attention have been given to manage traffic effectively. Vehicle License Plate Recognition based on image processing and pattern recognition is the key technology in intelligent transportation system. It can be widely used in road pricing, traffic regulation,off-site law enforcement and other fields. And it can raise working efficiency,save labor costs, and have great practical significance and economic values. Vehicle License Plate Recognition consists mainly of license plate location, character segmentation and character recognition. This paper discussed the three parts, proposes new algorithm as to the previous algorithm, and design algorithms in GPU and CPU platform.Finally give a software of Vehicle Plate Recognition based on CPU and based on CPU and GPU.(1). License plate location, this part analyzes the license plate location algorithm based on edge and MSER. For the low speed and poor accuracy of algorithm, three-valued feature images and License Plate Location Algorithm is proposed. The algorithm is optimized in the CPU and GPU platform.It has been proven that the algorithm have high accuracy and speed.(2).Character segmentation, this part discusses the segmentation algorithm based on projection. For the poor accuracy of segmentation, a character recognition based on character features and 3-value image and segmentation of multi decision is presented, and designed in the GPU and CPU platform. It has been proven that it can adapt most of the segmentation and algorithm has high accuracy.(3). Character recognition, this part discusses the Template matching method, for the poor accuracy of recognition, a multi-character library method of the statistical characteristics is proposed. It can directly reflect the characteristics, it has been proven that algorithm has high accuracy and fast in training characteristic databases.(4).Software design, A Vehicle License Plate Recognition software is given with C language, and a large number of License plates have been tested. After thousands of pictures testing, it shows that the recognition rate of the last six characters reaches at 99.2%,and of total plate reaches at 97.4%.
Keywords/Search Tags:license plate location, three-value image, GPU and CPU, character segmentation, character recognition, multiple decision
PDF Full Text Request
Related items