Study On Crack Detection And Identification Technology Of Straddle Monorail Traffic PC Beam Surface

With the acceleration of China’s rapid economic development and urbanconstruction, increasing pressure on urban traffic, the development of rail transportationhas become a mainstream mitigation of urban traffic pressure.PC track beam is animportant component of monorail transport system,its health status has a verysignificant impact on the safe operation of the entire system.Therefore, safety testingand conservation is crucial to improve the service life and security of the system.Thetraditional manual detection method has the following disadvantages,such as lowefficiency, high costs, poor accuracy,and so on.This paper researched many aspects ofthe beam surface detection system.This paper first analyzes the framework of the PC track beam surface crackdetection system, described the working principle and system design details.The wholesystem is divided into two major parts,including the off-line image processing systemand the image acquisition system.In the image acquisition system, describes the designparameters and characteristics of the various subsystems.In the image processingsystem,firstly, according to the characteristics of the beam crack imageanalysis,proposed an image pre-processing method based on wavelet analysis and fuzzytheory.After wavelet analysis, the low frequency coefficients of the image are enhancedby fuzzy function improved, and the high frequency coefficients de-noising by waveletthreshold function.This method of beam surface image preprocessing has goodresult,and a higher signal-noise ratio,the visual effect and contrast of the image has beengreatly improved.Secondly, using the improved Otsu-genetic algorithm for imagesegmentation.The traditional Otsu method is only suitable for the bimodal histogramimages,and the local threshold method has large amount of computation.This algorithmovercomes these shortcomings and has some advantages,such as computingfast,accurate segmentation,and so on.Use a algorithm of growth along the Consistentdirection of path, successfully eliminates most false crack and create favorableconditions for the identification and classification of crack image.Finally,design a threelayer BP neural network for crack identification and classification.The image projectionpeak ratio,the distribution density of crack D and the crack area pixel number sum areselected to carry on the training of neural network and accurate classification of imagedata.the error is≤5%,which in the range of system requirements. So the results verifies the validity of the method.According to the design requirements.the research results of the PC beam crackdetection system in this paper can detected beam surface cracks.It provides the databasis for the repair and maintenance of the rail system,has some reference value.