| With the development of transport, there are more and more vehicles, the management of the vehicle is also more difficult. IC card is regarded as a unique identifier in most paid parking. Once the IC card is lost, that will spend a lot of manpower to query the parked information. The introduction of license plate recognition technology in the parking lot system, extracting vehicle license number as the vehicle ID, through that you can make more intelligent management, reduce labor costs, will also make a significant economic value. This topic is to develop a license plate recognition system that based on the advanced recognition algorithms, which is based on a large car park management processes in Dalian, The development of license plate recognition system will improve the management level of the parking lot and the regulatory capacity of the vehicle.This research and implementation of license plate recognition algorithm is divided into five steps:image preprocessing, license plate location, tilt correction, character segmentation and character recognition. The focus of this research is license plate location and character segmentation. The purpose of image pre-processing is to highlight the license plate features, make it easy to positioning plates. In the positioning algorithm, taking the advantages of projection positioning algorithm and proposing a algorithm which is based on the combination of projection and edge distance, compared to ordinary projector positioning method, the proposed algorithm is more accurate, The accurate positioning can be more than96%for various models. In the tilt correction algorithm, to obtain the gray plate tilt angle by calculating the center of gravity, and then using an improved manner of rotate the image, while the horizontal and vertical tilt can be corrected. In the character segmentation algorithm, a segmentation algorithm is proposed which is based on clustering and projection analysis, it not only can split the normal character and to the fracture and adhesion character also have a better segmentation, the segmentation accuracy rate can be98%. In the character recognition part, first regulated the width and height of the character by the normalized of the character, and reducing characters’redundant by refining the character, last introducing the modified Hausdorff distance, using template matching method to identify the characters. Finally, the paper introduces the subject of the other modules of the system and key technologies. By querying client, this paper designed an algorithm that can efficiently solve similar character template matching algorithm which will lead confusion.Specific implementation of the algorithm is based on Visual Studio2008, using C++language. Laboratory and field tests showed that the software which is designed by the proposed algorithm when used to identify the low-resolution image, the recognition rate can over95%in a good weather condition, algorithm and software all perform excellently. This algorithm will have a considerable practical value and promotion prospects in the stop sign field. |
