Complex Background Of The License Plate Location And Character Segmentation

The research of special object recognition under complex conditions is a hotspot in IT and is a fairly active field, which is continuously enriched itself by adopting the new concepts and new methodologies from related fields and is of great importance in daily life and national defence, military affairs. As a special case of the special object recognition under complex conditions, VLPR (Vehicle License Plate Recognition) is presented in the paper. The main contents are two aspects: Object location, Object segmentation.The contrast of the vehicle license plate given by the national vehicle license plate standard is specially designed with largest contrast (four colors combination with largest contrast: black characters, yellow background and black border line; white characters, blue background and white border line; red or black characters, white background and black border line; white characters, black background and white border line) for easy visual recognition. Because of the weather, illumination and other factors, the collected license plate images contrast is not desirable. Especially the vehicle image's strong contrast disturbance made by the vehicle forward light in the night and the sunlight in the summer is inevitable.The vehicle license plate is composed of characters, which is edge in the binarization image; the color couple of the background and character are fixed in the license plate in the mainland; based on these, an all-weather, rapid, stable locating approach is presented in this paper. The paper presents a gray image enhancement algorithm; based on the color features of vehicle license plate, the definition: block hue average (BHA) is given. Firstly, enhance the collected image grayed, compute the BHA of the vehicle image collected in the real situation and eliminate the more non-vehicle-license-plate blocks. Then detect the edges of the enhanced image and the image processed by BHA, remove most of the background and noise edges based on the threshold of the lengths of the edges. Finally, search the plate in the residual edge image through projection operation and segment it out from the original vehicle image. And rectify the slanted and distorted plate image; also remove the four border lines.Compute the height and length of the vehicle license plate in the binary image based on a reliable binary threshold selection formula, and segment each character based on the feature of character-interval-character in the plate, through the horizontal...