Study On Intelligent Vision Measurement Technology For Planar Sheet Metal Parts

Sheet metal working is widely used in automotive,equipment manufacturing,electronic communications,home appliances and other industries.The mass production mode of modern manufacturing raises higher demands on measuring speed and accuracy of sheet metal products.This paper focuses on the application of vision measuring technology in planar sheet metal measurement.Several key technologies are discussed and implemented,including camera calibration,image edge detection,image mosaic,image registration and graphic primitive vectorization technology.The main contributions are as follows.On the study of linear and nonlinear camera calibration model,a four-parameter model with second-order radial distortion and first-order decentering distortion is proposed,as well as the solution method based on two-step iteration.By applying distortion parameters derived from the model to image distortion correction,the precision is compared with that of Tsai’s two-step method.In addition,the correction method of the image perspective deformation is discussed.Based on single pixel width image contour extracting using Canny operator,the contour is expressed by Freeman chain code.To improve contour location precision,the performance of three subpixel contour location methods is compared,including the curve fitting method,gray moment method and Zernike moment method,and the subpixel location is realized by Gaussian fitting method.The DXF(Drawing Exchange File)extraction process is studied for DXF-based automatic measurement,including the extractions of graphic primitives and annotations.Then registration method of DXF image with sheet metal image is studied,on the basis of rough registration by geometric characteristic method,ICP algorithm is adapted to realize fine registration.Considering that large sheet metal is difficult to be imaged one time,the image mosaic is completed by extracting SIFT feature point and the mosaic precision is evaluated.Intelligent graphic primitives recognition is carried out based on contour feature point extraction.After extracting circular contours through existence probability map,the feature points of complex contours are obtained by SVM method.The contour between two feature points is considered to be a basic graphic element,an integrated recognition method based on fitting error is used to judge contour segments type.To ensure the closure of parameterized contour,a twice traversal method is proposed.In the fitting process of graphic primitives,the constraint condition is established according to the positional relation between graphic primitives.The hardware and software of measuring platform are designed,including the selections for vision system and motion hardware,the overall software architecture,the measurement process and the main software interface.Based on intelligent recognization of graphic primitives,the measuring precisions of complex contour and circular contour are evaluated.