Research On Fault Image Detection Algorithm Of Center Plate Bolts And Brake Shoe Key

The railway freight transport in China has the characteristics of heavy load,heavy traffic density and complicated operating environment.Therefore,higher requirements for the safe operation of freight cars are put forward.TFDS(Trouble of moving Freight car Detection System)which can detect dynamically vehicles on the move,can detect and troubleshoot the on-line trouble in railway freight,and provide effective guarantee for the safety of railway freight.However,at this stage TFDS is still in a train inspection man-machine combination,it relies on human vision to judge the fault image to complete the freight car detection.The automatic recognition of the fault image by computer is the inevitable trend of the development of TFDS.The realization of this technology will quickly improve the measurement capability of railway freight safety inspection,and ensure the safe operation of freight cars more effectively.The fault images of center plate bolts and brake shoe key in TFDS system are chosen as the research objects in this paper.The related theories of machine vision,image processing and machine learning are used to design and implement the corresponding automatic detection algorithms.Aiming at the detection of fault images of center plate bolts,the fault images are classified,and sumed up similarities and differences,then the location and recognition methods of the fault images are designed.In the aspect of fault image positioning,the intersection points of the brake beam and the brake rod are located by using the common points of the image features.At the same time,the image types are identified by feature expression on one side of brake beam.As A type image,the area of reinforcing rib is located by the relation between intersection point and itself parameters.The exact location is obtained by the edge detection and projection.The location of center plate bolts is assigned finally through reinforcing rib.As B type image,open-sided area of center plate bolts image is processed by differential excitation from Weber Local Descriptor(WLD).Locate the longitudinal coordinates of the bolt by projection.And the Support Vector Machine(SVM)location classifier of the sliding window is used to detect the position of the bolt in this area.The other location of center plate bolts is assigned finally through the information just got.In the aspect of fault image recognition,the WLD-LPQ eigenvector of the bolt position,which is a combination of Weber Local Descriptor(WLD)and Local Phase Quantization(LPQ)is extracted,the fault recognition classifier is introduced into SVM to get the detection result.Therefore,the recognition rate of bolt failure has been improved.Aiming at the detection of fault images of brake shoe key,firstly,the brake beam is located,the possible area of the brake shoe key is determined,the suspected brazing areas of the brake shoe key are extracted,through the processing of a series of image processing techniques.Then the suspected region images are described with the WLD-LPQ feature vector,the SVM positioning classifier is introduced to locate the brazing area of the brake shoe key.The fault detection results of the brake shoe key are obtained by using the feature vector of the position of the brake shoe and introducing the SVM fault recognition classifier.Finally,the fault detection algorithm is optimized to improve the performance of the fault detection in this paper,and the latest data test algorithm is used to test the performance of the fault image,and its performance is analyzed.The test results show that the fault detection algorithm in this paper has good detection performance and can meet the actual requirements of TFDS automatic detection.