Research On Image Stitching Of Subway Bottom And Bolt Detection System For Key Components

With the development of urban rail train technology,the subway has become an indispensable means of transportation in people's daily life.In order to ensure people's travel safety,the safety of the subway cannot be ignored.The detection of the state of the bottom of the subway is one of the important links in the subway detection,but there will be missed inspections during segment inspection and factory inspection.At present,there is no complete set of subway bottom reproduction technology and image-based automatic detection technology in China.Therefore,this thesis has done a corresponding research on the bottom of the subway.The main work is as follows:(1)Design the overall architecture of the image stitching and key component bolt detection system at the bottom of the subway.Based on the needs analysis,determine the system architecture,system workflow,and component modules.(2)The horizontal and vertical stitching of the images taken by the line scan camera;first,the image is corrected and the data is cleaned.Due to the particularity of the line scan camera,it can be directly stitched vertically.For the horizontal stitching of the multi line scan camera,sift feature matching is used.And image fusion technology are used to stitch the subway bottom image,based on the traditional sift feature,the stitching efficiency is low,and the system has high real-time requirements,so this article first determines the overlapping area,and reduces the subsequent iteration time by performing feature matching on the overlapping area.To improve efficiency and complete the stitching of subway bottom images.(3)The traditional template matching and positioning technologies are not applicable to the subway bottom image.In this paper,the background difference is used to determine the position of the front and rear of the stitching image,and the position of key components is determined by proportional division.Modify the loss function to improve the convergence speed of the system model,analyze the yolo algorithm flow,and fine-grain the picture to improve the detection accuracy of the system to meet the metro detection accuracy requirements.(4)According to the system design,the system is installed and debugged at the Zhenlong depot of Guangzhou Metro.The feasibility of the algorithm is analyzed by using evaluation indicators.The experimental results prove that the algorithm meets the system requirements.