Research On Locomotive License Plate Character Segmentation And Recognition Algorithm

With the rapid development of the high-speed railway in our country,the number of trains is becoming larger and larger,and the management of the trains is also becoming more and more important.However,locomotive number as a locomotive's "ID card" can not be neglected in the management of locomotives.This paper focuses on the practical application of locomotive maintenance points,and studies the segmentation and identification of locomotive characters in locomotive number recognition system.It can identify the railway locomotives in and out of station immediately and meet the actual needs.In this paper,the status quo of regional segmentation and recognition of railway locomotive license plate number is introduced firstly.Based on the existing algorithm of car license plate character segmentation and recognition,the algorithm design and improvement are combined with the characteristics of railway locomotive license plate number.This paper mainly completes the image preprocessing after the locating of the locomotive number,the car character segmentation and the character recognition algorithm of the car number area.The main work of this paper is as follows:1.Preprocess the image of vehicle number region to remove the noise that affects the character segmentation and recognition,including the grayscale image of the region of the color number,and then use the method of grayscale stretching Image contrast enhancement.Binarize the image after contrast enhancement separates the target from the background to extract the effective area.After binarization also need to remove the noise and interference area.2.Design and implementation of character segmentation algorithm based on connected domain and improved projection-based character segmentation algorithm.Among them,the improved projection segmentation method combines the characteristics of locomotive car number segmentation.Experimental results show that compared with the character segmentation method based on connected domain,this algorithm has better adaptability to the image with good contrast and better segmentation accuracy.3.Design and implementation of two methods of character recognition algorithm based on template matching and BP artificial neural network.The experimental results show that the traditional template matching method is too sensitive to noise,while the BP neural network is more adaptive to noise.After a certain number of training sets trained the recognition model,it has a higher recognition rate.