Pattern Recognition of Part and/or Workpiece for Automatic Setting in Production Processings

Abstract

Summary A method of pattern recognition of part and/or workpiece for automatic setting applicable to the processings such as assembling, machining, clamping, etc. is proposed. A prismatic or polyhedral object is put at the recognition station equipped with a mirror, and both direct and mirror images are taken into a single picture by an ITV camera. Based on the picture points, the 3-dimensional coordinates of feature points or vertices of the object can be calculated when the relative position and posture of both the camera and the mirror are known. By utilizing the calculated coordinates and the connective relations of feature points, the geometric information on the shape, dimensions, and location of the object are reconstructed, and recognition of a specific or instructed portion of the object is possible. In this paper, the principle for calculating the 3-dimensional coordinates of feature points from the 2-dimensional coordinates of picture points in the projected direct and mirror images is described. The experiments by means of semi-automatic processings, in which the 2-dimensional coordinates of feature points in a projected picture are measured by eye-measurements, yielded reasonable results. Furthermore, the methods for detecting the feature points, solving the correspondence problem of feature points in the direct and mirror images and determining the geometric structure of the object are discussed.