Key Technology Research On Detecting System For Identification Of Hot Heavy Rail Based On Machine Vision

With the fast development of national economy in China，railway technology isapproaching to the way of high-speed，large-capacity and automation traction. Heavyload and high speed for track is becoming an inevitable trend, which requires higherquality for heavy rail of railway. Part of heavy rail produced in China is exported toother countries now. The trademark of heavy rail among exported products would gettightened inspection. The one of important element affecting the quality of entire heavyrail is the quality of heavy rail trademark. The quality of trademark and its productiveefficiency would influence the whole economic benefits of heavy rail.This subject is proceeding with research under the background of quality detectionfor thermal state heavy rail surface in a beam rail mill of a certain large steel enterprise.In accordance with the statement that the mill is still depending on artificial detectivemethods to observe the quality of heavy rail, a set of mark detection system under thethermal state of heavy rail is designed. This system is using machine recognitiontechnology to carry on an intelligent detection to the marked recognition of heavy rail,which overcomes some weakness such as the low efficiency of artificial detectivemethods, the strong influence from subjective factors, the intense strength of labor forworkers, low timeliness for detection and a large number of blind area of detection andfinally realizes online real-time detection to the mark recognition of heavy rail. Thesystem enjoys a very high practical value in industrial nondestructive testing field.The system is adopting the technique based on machine vision to detect theheavy rail. Because of the special application environment and these influential factorssuch as low light, heavy dust and high noise in production spot, the vibration ofequipments and something like that, there are three key points in realizing process for it:acquisition of the original image information for the system, the preprocessingalgorithm research of images, character recognition technology.The designed system in this paper will carry on a detection to hot rolling markinformation of heavy rail under the condition of complicated lighting environment, so itasks a higher requirement for the acquisition technology of original information for thesystem. Acquiring high-quality original images information becomes an importantresearch.The purpose of images preprocessing is to detect the images and divide the mark information from the images. The paper takes the preprocessing process of images ofheavy rail as the main line. In this paper, the algorithm like the image enhancement,image segmentation, mark area positioning is studied. First of all, the images werecarried on prepr0cessing research aiming at the problems of uneven grey contrastexisting in images, images with speckles and unclear mark edges. And then, accordingto high-brightness linear characteristics in images of heavy rail, the background regionalsegmentation to images was finished. Finally, by the research to traditional imagesegmentation method，It was successful to put forward a method which was based onimage level and vertical projection was totally suitable to the system, and which carriedout made location and character segmentation to image ID region.The character recognition technology is the key points in the system. In the light ofthe quality of few mark samples，in the analysis of various classifier based on theadvantages and disadvantages，SVM for character recognition was picked. After successin processing of segmentation and normalization of the characters，the extractionmethod to features of recognized characters for heavy rail was researched. Later, usingthe features training support vector machine classifier extracted helps to finishsimulation of character recognition with MATLAB and LIBSVM toolbox.