Font Size: a A A

Measurement Of Crack Length Based On Monocular Vision Ranging

Posted on:2019-04-21Degree:MasterType:Thesis
Country:ChinaCandidate:N DingFull Text:PDF
GTID:2392330551960009Subject:Control Engineering
Abstract/Summary:PDF Full Text Request
In recent years,with the rapid development of computer image processing technology,more and more people apply this technology to crack measurement assessment of the building surface.Compared to traditional artificial visual and acoustic measurement methods which have many drawbacks such as low objectivity,low precision,low efficiency,high cost,image processing technology has advantages such as non-contact,low cost,convenient,and high precision.As the surface of the wall is subjected to long-term changes in temperature and humidity or external tension within the load-bearing structure,thus the wall surface cracks become increasingly obvious,the wall surface cracks affect the visual beauty;and can even lead to wall collapse.The current working and structure status can be reflected in the crack characteristics according to the crack characteristic.It has very important practical significance that judging the severity of crack according to the crack characteristic parameter values.Based on the multi-crack characteristics and environmental conditions of the wall,this paper studies and discusses the method of crack width measurement in wall crack image.The innovative opinions of this paper are as follows:(1)Aiming to the problem that net crack width of crisscross pattern is difficult to measure,we propose clustering analysis method to separating wall cracks;then through scale method,minimum distance method and tangent line three crack width calculation method is verified.Considering the crack type,measurement time and measurement accuracy,the vertical tangent method is proposed to measure the crack,and the average relative error is 9.61%.(2)To solve the problem that the measure err of single color bar is too big.We use standard color bar to compare the method of measurement.The measured range was 0.5 ~ 3mm.The average relative error is 6.42%,which is 34.62%lower than the average relative error of 9.82%.
Keywords/Search Tags:Crack detection, Image processing, K-means, Feature extraction, Monocular vision, Network crack
PDF Full Text Request
Related items