| With the development of economy, the car can’t meet the needs of transportation,thus train has been the mainstream in the transportation. But with the increasing transportation tracks and train numbers, the security has became a hot topic of concern. Recording routes and train tracks by hand is a waste of labour and financial resources. Due to the development of computer technology and image processing technology, designing a set of the carriage number of high-speed train automatic identification system by internet is the best choice to monitor real-time train running track. The accurate carriage number positioning and identification are the core of the problems for the automatic identification system. Because the train on the railway track is high-speed running, and the surrounding environment of railway is complex,so the greatest obstacle is to collect high quality original image. While carriage number is print fonts, according to the requirements of different people, the font, size and rules of the carriage number are not unified. Moreover the font of the carriage number and the train body part have the same color. And the collection is limited by camera hardware and other equipment. Both external factors and internal reasons are likely to make the original image fuzzy and gray. The original image with lower quality will directly induce errors of image positioning and character recognition.At first, this dissertation studies the status quo of the high-speed train position technology and identification system, accesses to the large number of domestic and foreign technical literature and data, analyzes location identification technology difficulties. Then there are a series of image processing for the original image, such as the size of normalization, luminance correction, gray, filtering, sharpening, swelling and corrosion etc, making the image edge feature information more clearly. Second,image edge detecting, image binary with threshold method. This dissertation puts forward a applicable to this thesis railway carriage number area two targeting methods of coarse location and accurate positioning based on the edge of the carriage number area information more than other regional characteristics. The coarse positioning method based on edge information locates the active position of thecarriage number area, eliminating the interference fringe information. The precise positioning method making use of the horizontal projection and vertical projection could accurately find out the upper and lower boundary line and area around the boundary line of carriage number. Then performing a single character segmentation and color fill for the localization of the target area image. And finally, image identification use three-layer BP neural network recognition method. Above algorithm compiled a set of software system is suitable for field operation with VC++platform and OpenCV library. Proved by experiments in this thesis, the carriage number of high-speed train automatic identification system has a good real-time performance and better accuracy. This method is applied to a passenger train carriage number identification system, and the test shows that this algorithm design is feasible. |
