Study On The Anomaly Detection System Of Fingerboard For Straddle Type Monorail Traffic Based On Line Structured Light 3D Vision

As a kind of electric energy-driven,fast and convenient public travel mode,straddletype monorail is favored by many cities,which is of great significance to solve the growing traffic congestion problems.The fingerboard is used as a connection between the two rail beams.The abnormality of the fingerboard will seriously threaten the smooth passage of the train.Therefore,the abnormal fingerboard will seriously threaten the safe operation of the train.At present,manual inspection is mainly used to detect abnormal fingerboard,which is inefficient and inaccurate.And no mature and efficient automatic detection system has been found at home and abroad.Therefore,it is of practical significance to realize the automatic detection system for the abnormal fingerboard of straddle-type monorail.This thesis proposes to use line structured light to realize the collection of 3D visual data of the fingerboard.The detection of abnormal looseness or shedding of fingerboard fasteners and misalignment is realized through 3D visual analysis.This thesis mainly covers the following contents:1、The fingerboard anomaly detection scheme based on line structured light 3D vision is proposed.According to the analysis of the research object and the on-site working environment,the lens,line structured light source,camera,encoder and other devices were selected.A straddle-type single-track 3D data acquisition system was built,and the geometric relationship and relative positions of the components in the acquisition system were determined to ensure the quality of the collected data.2、The preprocessing method for point cloud data of fingerboard is introduced,the noisy points are removed combining coarse filter and statistical filter.Based on the geometric features of fingerboard,two-pointer segmentation algorithm is proposed to segment the fingerboard.In the end,the voxel-based grid filter is adopted for down sampling,which makes the data more concise and improves the operation speed of the following algorithm.3、For the abnormal fastener of fingerboard,the target detection network of point cloud is used to locate the fastener area,and then the average height of the point cloud and the height of the datum in the fastener area are calculated respectively.This is analyzed to determine the fastener loosening or shedding and other abnormalities;For the fingerboard misalignment,the position of the fingerboard is obtained by locating the position of the fastener area and combining the geometric characteristics of the fingerboard.The misalignment is determined by comparing the height of the finger part of the two fingerboards.The experimental results of the simulation system built in the laboratory show that the anomaly detection system of fingerboard for straddle type monorail traffic based on line structured light measurement technology has a collection accuracy of ±0.1mm.The anomaly detection algorithm is carried out in the test picture.In the test,the accuracy of2mm-4mm misalignment detection reached 97%.For the abnormal detection of fasteners,the precision reached 96.4%,while the recall rate reached 98.5%,which proved the feasibility of this method.