Abstract: In order to guide the automatic alignment of axis and hole in the process of robot assembly, a high-precision visual detection algorithm of wafer pose is proposed. According to the geometric properties of the wafer, the contour of the wafer in the captured image is extracted, the upper surface of the wafer is detected and the sub-pixel edge point is searched, so as to achieve the high-precision fitting of the contour ellipse equation of the upper surface of the wafer. Based on the fitted ellipse equation and the perspective projection principle of conic curve, the mathematical model of pose transformation from 2D ellipse to 3D space circle is established, which can solve the center coordinates of the upper surface and the direction vector of the axis of the wafer according to the ellipse equation of the surface contour and the radius of the wafer. Two contour points' screening criteria are proposed to eliminate the interference of some non-Gaussian noise in the image. To deal with the ambiguity of pose solution, an ambiguity elimination method based on the reprojection of the circle center on the lower surface of the wafer is proposed. Its advantage is that it does not need additional environmental constraints. The simulation analysis and visual inspection experimental results show that the proposed algorithm has good robustness, and the detection accuracy is as follows: the detection accuracy of x and y axes of the center on wafer's upper surface is 0.01 mm, and the angle detection accuracy of the axis direction vector of the wafer is 0.7°.

Key words: pose detection, wafer, ellipse detection, sub pixel, ambiguity