Research On Feature Recognition Technology Of High-speed Railway Bridge Cracks Based On Image Analysis

In recent years,the networking effect of high-speed railway in China has become more and more prominent,and the maintenance task of railway bridges have also increased.Cracks are one of the most serious hidden dangers that threaten the safety of high-speed railway bridges in service.Therefore,it is of great practical significance to realize rapid and accurate detection and dimension measurement of cracks,and is also a key research topic in the field of high-speed railway bridges.At present,crack detection and dimension measurement of high-speed railway bridges in China mainly rely on manual inspection,which is not only time-consuming and labor-intensive,but also has limited detection accuracy.With the continuous practice of automation technology in the field of bridge crack detection,and the great advantages of deep learning technology in the field of related target detection,the use of automation technology to achieve automatic monitoring of high-speed railway bridge cracks has become the research consensus of scholars in this field.However,the existing crack detection and identification technology is far from meeting the needs of engineering practice.Therefore,the automated crack feature recognition technology needs to be studied urgently.In this paper,the crack image of high-speed railway concrete bridge is taken as the research object,and the crack classification,feature extraction and dimension measurement are carried out.The main research contents are as follows:(1)Pre-processing and classification of crack images.First of all,in order to obtain high contrast image,an improved Mask algorithm is proposed to solve the problem of uneven gray distribution of the original crack image,and to enhance the cracks feature by using the deblurring algorithm.Then an intelligent crack detection method based on improved Faster R-CNN+ZF model is designed,and making the crack data set input network for training,the automatic classification of all cracks is realized,and the overall classification accuracy reaches 93.7%,which provides a new classification method for crack detection of high-speed railway bridges..(2)Feature extraction of cracks.For the crack-containing images selected by the classification,the median filtering algorithm of the sparse template is first used for image denoising,and then the improved perspective transformation method based on four-point mark is used to obtain the standard front view correction image.Finally,high quality crack features are extracted by threshold segmentation,secondary connected domain denoising,edge detection and morphological processing.(3)Measurement of crack length and width dimensions.Firstly,the camera's internal,external and distortion parameters are obtained by camera calibration technology.Then a piece wise polynomial fitting algorithm based on feature nodes and a piece wise fairing curve splicing algorithm based on boundary points are proposed,by fitting a single-pixel wide centerline of the crack profile feature,the non-contact measurement of crack length is achieved.Then a crack width measurement algorithm based on single pixel dimension and a quantitative corrosion algorithm are proposed to achieve the average width and maximum width measurement of crack features.Finally,the actual sample is selected for dimensional measurement test,and the measurement accuracy of each parameter meets the requirements of engineering practice,this proves the robustness and practicability of the proposed algorithm,and can realize the automatic measurement of cracks in high-speed railway bridges,which provides a more effective means for the measurement of cracks in high-speed railway bridges.