Research Of License Plate Recognition System Based On DSP

The License Plate Recognition System is an important part of Intelligent TransportSystem, which is a research in the field of computer vision and pattern recognition. TheLicense Plate Recognition System plays an important role in regulating the transportation.But the realization of the system is very complexity, some problems should be solvedurgently and more research should be paid to this technique is very necessary.This thesis makes a research and discusses about the key and difficult points indesigning and implementing the embedded automatic license plate recognition system,and design a LPR hardware platform based on the newest DSP kernel C64x+with theideas and principle of chip selection, circuit design and driver software design. The threemain modules of the platform is the image acquisition module, the memory module andthe Ethernet interface module. All the modules are proved to be work well throughsoftware test. In the process of designing and implementing the software of the system,this thesis designs a method combining the traditional license plate image preprocessingsteps and mathematical morphology thought in the stage of license plate imagepreprocessing, which reduce more image noise and interference to the plate informationwith experimental results. In the license plate location stage, this thesis combines thededicated video processing functions on the embedded platform with a license platelocation method using image binaryzation and puts forward a new method of vehiclelicense plate location which is proved to be validated through experiments. Finally, thisthesis expounds the license plate character segmentation method based on value ofprojection has good performance on this platform and from the overall elaborates thisplatform base on DM6437DSP can realize the basic function of the LPR system.The system designed and implemented in this thesis accomplishes the function of thethree modules: license plate preprocessing, license plate location and charactersegmentation. The stability and real-time of the system reach the expected goal, thatprovide a train of thought for the embedded LRP system.