Inspection And Evaluation System For The Shape Of Steel Weld Based On Multi-Line Structured Light

Steel bars are commonly used in construction,bridges and other projects,and their welding quality directly affects the service life of the project.The strict inspection of the welding quality of steel bars is an indispensable part before pouring and use,which is of great significance to ensure the safety of the main body of the building and the safety of people ’s lives and properties.In response to the current status of China’s traditional manual inspection and the related inspection system with insufficient integrity of detection indicators,low inspection efficiency,inconvenient operation and subjective quality evaluation mode,a reinforcing steel weld shape detection and evaluation system based on multi-line structure light is designed.The thesis takes the lap-formed steel weld as the research object,and proposes and implements a multi-line structured light-based steel weld shape detection and evaluation system by studying multi-line structured light 3D detection technology,visual sensor calibration technology,multi-line structured light stripe center extraction on the weld surface,weld shape feature detection algorithm,quality evaluation method and evaluation software design,which is of great significance to the detection and evaluation of the appearance quality of steel weld.In the study of the inspection and evaluation of the shape of steel weld,the principle of multi-line structured light visual inspection is firstly described,and the process of establishing the inspection model of the system is given.The mathematical model of the multi-line structured light visual inspection system is derived,the calibration of the visual inspection system is completed,and the mapping relationship between the image coordinates of the weld to be measured and the world coordinates is established.Next,a handheld steel weld image acquisition instrument is designed.Using the FPGA core processing unit,the design of weld image acquisition,pre-processing,caching and transmission are completed to lay the foundation for the subsequent detection of weld’s morphological features.Then,for the difficult problem of extracting the center of multi-line structured light stripes on non-uniform reflective surfaces,a multi-line laser stripe center extraction method combining U-Net depth network and Steger algorithm are proposed.The experimental comparison results show that the method can effectively solve the problem of light stripe image processing on non-uniform reflective surfaces.Then,the weld width,residual height,thickness,saturation and porosity feature detection algorithm are designed to complete the detection of weld forming quality,and a multiindicator evaluation model is constructed using information entropy as the evaluation index to achieve objective judging of weld appearance quality.Finally,the system hardware and software platform are completed,and the feasibility,accuracy and stability of the system are verified by conducting experiments on parameters such as weld thickness,width and residual height,and comparing the results with those of precision instruments such as vernier calipers and weld inspection rulers.