Machine Vision-based Bulk Angle Identification of Loose Steel Structure Bolt Groups
In steel structure buildings,the loosening detection of nodal bolts helps to ensure the stability and safety of the steel structures.In order to solve the current problems of bolt loosening in terms of detection angle range and batch detection limitation,a machine vision-based automated batch angle recognition method for steel structure bolt loosening is proposed.Firstly,the target frame of the bolt group area is located by the target detection network.The extracted bolt area is calibrated and numbered according to the extracted bolt area.The combined processing flow of mask construction,ROI segmentation and duty cycle judgement is proposed.The pgonCorners angle point algorithm is used to extract the angle point location coordinates of the bolt group.Further combined with the color matching algorithm,the drawing and extraction of the corresponding index numbers of the marked point positions before and after the bolt loosening are completed.And calculation of the bolt loosening group angle is completed according to the coordinate system transformation algorithm and the array operation algorithm.The test results show that the method can accurately extract bolt-related features,the maximum relative error is 5.8％,and it can achieve accurate and fast detection of bolt loosening angle.
NETL
NSTL
万方数据
