|
The Vehicle License Plate Recognition is a critical research topic for modernintelligent transportation systems and an essential portion to achieve intelligence trafficmanagement. On the basis of analysis typical License Plate Recognition algorithm inrecent years, improves the preprocessing algorithm project based on using image detectiontechnology, lucubrates the troubles of license recognition under severe environment,especially (including plate stained on purpose or not, bad weather, slanting pictures byshooting angle, etc.). Pre-processing images of license plate, locating plates, segmentingcharacters and recognizing characters are the four manly modules in this study. Forpre-processing module, this study employs some methods including image correction,gray stretch and black-white enhancement, smooth denoising basing on median filtering,edge detection and image binaryzation using to obtain optimal threshold. In the licenseplate positioning pattern, this paper adopts the secondary precise positioning that bases onmorphological positioning, edge detection and vertical projection. In the license platesegmentation portion, this treatise applies license plate coarse segmentation, and verticaland horizontal projection of the license plate accurate segmentation algorithm. For licenseplate character recognition module, this thesis uses two algorithms for template matchingand neural network, and makes a comparative analysis. In addition, I set up an enormoussample library, and made lots of analogs under various kinds of server conditions thatcould interfere the recognition of characters, in order to obtain profound database to study.According to a great number of relative experiments, the emulation result shows utilizesneural network algorithm the recognition rate can be greatly improved.
|