A Method For Detecting Wall Defects Based On Binocular Vision

With the increasing aging population and rising labor costs,it is difficult for the construction industry,which is characterized by labor intensity,high work intensity,and poor environment,to recruit suitable workers.For example,in wall polishing,workers need to hold a handheld grinder flat or high to polish every corner of the wall,making it difficult for their arms to support for a long time,and the dust generated by polishing can also pose a threat to human health.This article takes construction robots as the research object and proposes a projection assisted binocular visual wall quality detection method for defect recognition during the wall polishing process.The main research content of this article is as follows:(1)Introduced the conversion relationship between the world coordinate system and the image coordinate system based on the camera imaging model;Introduced the principle and calibration method of binocular vision ranging;Analyzed traditional stereo matching algorithms and their accuracy;Analyzed the effectiveness of wall defect detection using various methods under ordinary light sources.Based on the data of laser dot matrix scanning total station,the deviation nephogram of binocular visual inspection results is obtained through coordinate conversion,so as to obtain the accuracy and limitations of each algorithm for wall defect detection under common light source.(2)To address the issue of low visual detection accuracy caused by the lack of feature points on the wall,a projection random grayscale image is introduced to increase feature points,and scanning line optimization and sub pixel fitting are used to improve accuracy;Based on the principles of sparse matching and differential evolution,a method for projecting scanned images,checkerboard grids,and dot matrix images is proposed to analyze the impact of various algorithms on robustness,time consumption,and accuracy.Research has shown that projection assistance can significantly improve the accuracy of wall defect detection.(3)Build three sets of devices: a simulated wall,a polishing device,and a visual device,and obtain the feed rate of the polishing device and the actual wall polishing amount through binocular vision method to verify the timeliness and accuracy of the proposed binocular vision algorithm.Through this study,it has been found that the methods of projecting random grayscale images,checkerboard grid images,and scanning line images can effectively identify wall defects.The NCC algorithm that projects random grayscale images and optimizes scanning lines and fits sub pixels can achieve a detection pass rate of 97.1%,with a detection error basically guaranteed to be within 0.7mm.The projection checkerboard grid method can achieve a pass rate of 98.6%,with a detection error basically within 0.5mm,and the projection scanning line method can achieve a pass rate of 98.6%,The detection error is basically within 0.4mm.Therefore,the proposed wall defect target detection method based on binocular vision has good application prospects and engineering value in areas such as building wall quality detection.