Research On Intelligent Inspection System Of Rail Laser Quenching Quality

The rapid development of railway transportation in China brought a great challenge to rail maintenance and repairing technology.The laser quenching technology can strengthen the surface of rail,extend the service life of the railway,and improve the safety of running train.In order to promote the industrial application of the rail laser quenching technology,our research group has designed a set of rail laser flying quenching system which includes measurement,processing and inspection module.This paper focuses on the research of the rail quenching quality intelligent inspection software.The details are as follows:(1)The intelligent inspection system for rail quenching quality has been designed and built.In terms of hardware,the light source,illumination method,camera and lens were analyzed and selected according to the requirements of machine vision inspection of rail surface.In terms of software,Visual Studio 2010 and MFC framework are used for software interface development.Based on the OpenCV visual library and the SDK development kit of the Vision CCD camera,the structure of intelligent inspection software and the algorithms for processing the rail quenching image were designed.(2)The image preprocessing algorithms for rail surface quenching were deeply studied.Firstly,the localization and grayscale processing of the rail quenching image was carried out to remove the non-rail area in the image and improve the processing speed.Then,the three filtering algorithms of mean filtering,Gaussian filtering,and median filtering were compared,and the median filtering algorithm of 5×5 template size was selected to filter the image noise.The histogram equalization method was used to enhance the image contrast.Then the local mean adaptive threshold method with the neighborhood size of 31×31,was used to achieve accurate segmentation between the background area of the rail and the target area of the quenching spot.Finally,combined with the morphological closing operation,a binary image of the rail with smooth regional boundaries was obtained.(3)The feature extraction algorithms of laser quenching spot on the rail surface has been studied deeply.Firstly,a quenching spot contour threshold selection algorithm based on the Canny operator was proposed to detect the edge of the rail binary image,extract the contour and image of quenching spot accurately,and calculate the shape feature parameters of quenching spot,so as to realize the accurate positioning of quenching spot.Then the gray level co-occurrence matrix method was used to analyze the grayscale feature and the texture feature of quenching spots,and the texture feature parameters such as energy,random entropy and contrast were obtained.Finally,the feature extraction experiments were carried out for laser quenching spots of three different energy levels.The various types of feature parameters of quenching spots were extracted and the feature database of laser quenching spot was constructed.(4)The intelligent classification algorithms for the quality of laser quenching spots on the rail surface were further studied.The CART decision tree model was selected,and eight characteristic parameters such as shape feature,texture feature and grayscale feature in the quenching spot feature database,were used to construct the data sample set for training.The comprehensive accuracy rate of the model is 96.7%.The support vector machine(SVM)model was constructed by linear kernel function,and the image features in the quenching spot feature database were used as image sample set.The SVM model with penalty factor of 0.1 has been trained,and the comprehensive accuracy rate of the model is 98.7%.