Research And Verification Of Image Acquisition Equipment And Analysis Algorithm For The Appearance Size Of Prefabricated Bridge Columns

Prefabricated bridges form the main body of the bridge by transporting different bridge components processed and formed in the prefabricated component factory to the site for lifting and splicing.The quality control and inspection of the appearance dimensions of prefabricated bridge components is a key link to ensure on-site construction efficiency.At present,the inspection of the appearance and size of prefabricated components mainly uses manual tape measurement,which has problems such as long inspection time,low efficiency,and poor repeatability of inspection results.Therefore,it is difficult to effectively control the external dimensions of prefabricated components,which seriously affects the efficiency of on-site assembly.This article focuses on quality control parameters such as the appearance size of prefabricated bridge columns,the spacing between spliced steel bars,and the elongation of steel bars.An image acquisition device and appearance size image analysis algorithm based on 2D and 3D image analysis technology have been developed to verify the detection accuracy.Firstly,this article compares and analyzes the advantages and disadvantages of the main2 D image acquisition devices.Based on the working principles of different types of cameras,considering equipment costs and the on-site environment and working conditions of prefabricated component factories,a 2D image acquisition device platform with a black and white array industrial camera as the core was built to collect images of the column end.Through on-site image acquisition analysis,the reliability of the built acquisition platform was verified.Secondly,based on a large number of 2D images of columns collected,an algorithm for measuring the distance between embedded steel bars of columns was developed.Using the adapthisteq function in MATLAB,the adaptive histogram equalization processing of the column end face image with limited contrast was realized,which solved the problem of image oversaturation.Gaussian filter was used to denoise and smooth the image.Finally,Canny operator was used to achieve accurate segmentation of the steel bars in the column image.This article also uses deep learning for pre embedded steel bar model training to improve the recognition effect of the steel bar part in the end face image of the column.In order to verify the accuracy of the image analysis algorithm,this article compared the calculation results of steel bar spacing based on 2D image analysis technology with manual measurement results at39 steel bar points.It was found that the average error between the calculation results of pure image processing technology and manual measurement results was 3.23 mm,with an average error rate of 1.61%;The average error between the detection results trained on the deep learning model and the manual measurement results is 2.08 mm,with an average error rate of1.056%.The accuracy of the two different analysis algorithms was compared and verified.Then,in order to accurately detect the elongation of pre embedded steel bars in prefabricated bridge columns,this article developed a point cloud data collection device for the appearance of prefabricated bridge columns based on Li DAR.The working principles of common different types of Li DAR were elaborated in detail.Based on the actual detection accuracy requirements and working environment,a single line To F Li DAR was selected as the point cloud data collection device.This paper specifically explains the design idea and acquisition process of the acquisition equipment.After the point cloud data acquisition on the end face and side face of the column on site,the point cloud data visualization on the end face and side face of the prefabricated bridge column is carried out,thus ensuring the effectiveness of the acquisition equipment.Finally,based on the collected 3D point cloud data,this article developed an analysis algorithm for the appearance size,embedded steel bar spacing,and steel bar elongation of prefabricated bridge columns.Through the voxel Downsampling algorithm,the number of point clouds is reduced while maintaining the integrity of the column point cloud data structure;Furthermore,the RANSAC algorithm was used to segment the point cloud data of the column end face(including embedded steel bars);Using DBSCAN clustering,the minimum distance value of algorithm parameters ε When set to 0.4,the filtering of point cloud noise points and outliers on the end and side of the column was achieved,achieving good point cloud clustering effect;Finally,using the Graham scanning method in the convex hull algorithm,the column end face,steel bar end plane,and column side plane were accurately segmented.Then,the length,width,height,spacing distance,and elongation of the column were calculated using the Euclidean distance formula.By comparing the detection results of the collection equipment placed at four measuring points 3m,5m,7m,and 9m away from the end face of the column with the results of manual tape measurement,it was found that the maximum difference between the two at the 5m measuring point was 4mm,and the average error was only 2.37 mm.The measurement effect was the best,verifying the accuracy of the measurement algorithm;By comparing the average errors of each of the four measurement points,it was found that as the distance increased,the error with the manual measurement results gradually increased.By placing the acquisition equipment on the side of the column and comparing the detection results of three measuring points(the distance is 6 m)at 30 °,45 ° and 60 ° with the column,it is found that when the angle is 45 °,the difference between the column height calculation results and the manual detection results is the smallest,3 mm,which further verifies the accuracy of the size measurement algorithm.