Crossbar Quality Detection System Based On Machine Vision

The crossbar of the truck is one of the important parts of the railway truck,its parameter indicators directly affect the safety performance of the truck,so regular maintenance and inspection become more important for the truck.Currently,manual testing is the main testing method,which needs a vast number of costs of labours and resources.Moreover,the workers should be technical and experienced.So manual testing doesn't satisfy the pursuit of efficiency for modern enterprises.At the same time,automatic detection of crossbars is still at the starting stage,its accuracy and convenience are far from the requirements of industrial testing.Aiming at above target of detection,the paper proposes the new method based on machine vision to measure the relevant parameters of the crossbar.The main content of the paper is shown below1.The binarization arithmetics based on adaptive thresholding is proposed.For the complex and changeable construction sites,the light differences between rain and shine or day and evening affect the accuracy of detection.By optimizing the image recognition process and using particle swarms optimization,the binarization arithmetics based on adaptive thresholding is designed to optimize the algorithm and reduce the effect of light changes on the picture extraction.2.The detection table detection algorithm based on projection mapping is proposed.For the certain error in automatic measurement,the detection table detection algorithm is explored.Firstly,the standard rod is used,and the mechanical engineer is required to calibrate and form a mapping table through the one-to-one positioning principle according to the requirements.In the actual measurement,the principle of regional positioning can reduce errors and improve detection accuracy,furtherly improves the effectiveness of the crossbar quality inspection system.3.The crossbar quality inspection system is designed.The system is divided into four parts:(1)Calibration of the camera.Through simulation and field test,the parameters of the camera are obtained,and the optimal placement position is selected;(2)Foreground target extraction.According to the scene environment,the paper uses the background difference method to process the captured image data and obtain the foreground target based on the principle of machine vision.The average background difference method and binarization arithmetics based on adaptive thresholding are analyzed and compared,and accurate ROI picture data is obtained finally;(3)Thecalculation of relevant parameters.After the spatially correction of extracted target,the flatness and length parameters are obtained by the geometrical calculation.For the reference point data,the accuracy cannot be guaranteed due to the influence of systematic errors,so the data mapping method is designed to reduces the influence of system error and improves the accuracy;(4)The expert system is designed.The crossbars are divided into seven grades according to the expert judgment criteria:first-class excellent,second-class superior,first-class good,second-class good,first-class middle,second-class,middle-class,scrapped,which is convenient for the later maintenance and repair work,and has printing function to output inspection reports on the purpose of reducing the using difficulty and improve practicality.Finally,after the implement of the crossbar quality inspection system,the experts would be invited to conduct data detection of the crossbars taken out randomly for comparison tests,in order to verify the effectiveness of system depend on whether the test results are completely consistent.