| License Plate Recognition (LPS) is critical part of the intelligent transportation system. It can be used in many kinds of vehicle management occasions. Compared with traditional vehicle management methods, LPS greatly improves the efficiency of vehicle management, reduces manpower and material resource, makes vehicle management scientific and normative, and safeguard traffic security. So it has expansive application foreground. LPS generally includes three modules: license plate location, character segmentation and character recognition.Based on image processing, computer vision and pattern recognition, the paper conducts a full investigation into distortional license plate recognition system. Our work includes:(1) The license plate image is restricted greatly by illumination. The license plate blurs in the condition of sunny or sunless such as in cloudy, flog and night. Aiming at this problem, histogram equalization and gray stretch are used to enhance license plate image, and make the image clearer.(2)The result of binarization directly affects character segmentation and character recognition. Local dynamic threshold of Otsu algorithm is improved in this paper. Experimental result shows, that this method not only has strong anti-noise ability and reserving character stroke, but also satisfying real-time need.(3)The declining license plate makes character segment difficult. To correction license plate, Mathematical morphology is employed to simplify the information quantities in license plate, then Hough transform detects straight line to achieve its declining angle, thus geometry distortional license plate correction is accomplished.(4)Based on the natural characteristics of license plate, rough character segmentation is realized by scanning line by line, and precise scanning is accomplished by combining vertical projection with scanning back.(5)Research and discuss character recognition. |
PDF Full Text Request