Research Of Holes Position Measurement Method On Curved Surface Based On Computer Vision

As the surface modeling of automobile panel becoming more and more complex, agreat challenge has been brought to industrial inspection for its more and more highprecision requirements of the assembly precision. Automobile assembly accuracylargely depends on size, shape and position accuracy of the hole on the automobilepanel. It is often hardly to assemble during vehicle assembly due to the aperturedeviation or position relationship of holes on the automobile panel. Thus the size andposition accuracy of the hole directly affects the assembly precision of automobilepanel. The hole of curved surface is detected mainly through the correspondingdesigned fixture. Each automobile panel needs to customize the corresponding fixture,which costs highly and low efficiency. So a set of detection method with highefficiency, low cost and generality is urgently needed in the detection of curvedsurface hole of automobile panel. To solve the problem, a3D measurement techniquebased on line structured light to detect the size and position relationship of holes onthe curved surface is put forward. Firstly, a set of hand-held3D laser scanning systemis developed to get the point cloud data of curved surface of the hole. Secondly, thepoint cloud of the hole boundary can be extracted by using the cloud boundaryinteraction extraction technology. After get the point cloud of the hole boundary, thesize and location of the hole is detected by the hole detection algorithm. This paperfocuses on the study of the three aspects.In terms of the development of Hand-held3D laser scanning system, the mainlystudy focus on the extraction method of the laser center and the registration of pointcloud based on fiducial marker. A method to extract center of laser by adaptive laserstripe width with direction template is proposed. The recognition criterion of theround fiducial marker is proposed according to the shape, dimension and graycharacteristic to recognize the fiducial marker. The algorithm of finding fourcorresponding common fiducial markers in different view is proposed to achievereal-time registration of the3D laser data. The accuracy of the3D scanning system is0.364mm. In terms of extraction method of point cloud hole boundary, the algorithm of interactive extraction of hole boundary on the scattered point cloud can fastly andaccurately extract the designated hole edge of the whole point cloud. In terms ofdetection algorithm of round hole, the detection method of hole on curved surface isproposed based on the principle that the round hole on the curved surface projects tothe vertical plane of the axial must be a circle. The precision of detection algorithm ofround hole is tested by the standard aperture of round hole on the CAD curvedsurface. Experimental results show that the hole axial vector measurement precisionis±0.0931°, radius measurement precision is±0.002mm. The position relation ofholes on curved surface of automobile panel can be accurately detected by the actualmeasurement.