Research On Truck Fault Image Recognition Algorithm Based On Machine Vision

With the rapid development of heavy-haul transportation in China,higher requirements have been put forward for the freight load,single-vehicle load,mileage and speed of railway freight trains.How to solve the safety problems that may be encountered in the process of transporting goods is a major difficulty at present stage.The trend of automation,because of its saving people,things,financial resources and high reliability,large amount of calculation and many other advantages,in all aspects of the industrial society flourish branch,industrial automation and has become a general direction.Therefore,automatic detection of railway freight car safety problems and automatic identification of faults have become a necessity.In this paper,based on the TFDS truck safety detection system the automatic identification of several faults that often occur in the operation of railway freight cars is studied.The main work is summarized as follows:(1)The possible image processing algorithms used in TFDS fault diagnosis are analyzed and compared,and the ones with better processing performance are selected to prepare for subsequent fault diagnosis.(2)A deflection fault identification algorithm based on edge deflection Angle detection was designed for the deflection fault of locking plate.According to the morphological characteristics of the target after deflection,the algorithm selects the Sobel operator to screen the edge in a specific direction,performs hough line detection on the feature edge,and performs automatic fault diagnosis according to its deflection Angle.The interface of automatic detection system is designed for single locking plate fault diagnosis and batch picture fault diagnosis.(3)A fault classification algorithm based on linear feature is designed for cross bar bending faults.According to the morphological segmentation characteristics of the bottom image of the cross bar,a method of locating the cross bar based on gray projection curve was designed to realize the location and extraction of the cross bar region.According to the fault characteristics of cross bar,choose Canny edge detection method,extract the edge of the crossbar fault characteristics,morphological characteristics of its processing enhancement,combined with the actual fault experience,design the failure criterion,adopted based on the least risk bayes algorithm to select the optimal threshold,and crossbar automatic detection system for single and bulk interface.(4)A classification algorithm based on HOG+SVM is designed for blocking key.According to the structure of the retaining key in the bogie,a positioning method based on mathematical model is designed and the retaining key is located successfully.The HOG feature of the gradient histogram of the block key region was extracted as the fault feature,and the SVM classifier was trained to achieve the classification and prediction of the block key fault,and the interface writing was completed.(5)The feasibility of each fault diagnosis algorithm is verified by experiments.The designed fault diagnosis system can realize automatic recognition of three kinds of faults,such as loosening of locking plate,bending deformation of cross bar and loss of retaining key,and meet the safety requirements of automatic diagnosis of TFDS diagnosis system.