| 摘要 |
A system for learning a characterization of an object and thereafter identifying the presence and orientation of the object is capable of controlling a robotic system for grasping the object. In an off-line learning mode, a television camera produces a video signal from the object and a preprocessor develops a segmented line outline of the object in digital form. A computer is programmed to generate a mathematical "rubber band" or convex hull around the object outline formed from segmented lines and to store the x-y coordinates of the lines. The centroid of the hull is calculated and an R(O) function as the radius from the centroid to the hull with the maxima thereof is calculated for characterizing the orientation of the object. A truncated Fourier series of the R(O) function is generated and the coefficients stored. In an on-line characterization mode, the apparatus generates a characterization of an unknown object and compares it with the learned and stored values of R(O) and Fourier coefficients. Upon identification of the object, the orientation is calculated by comparing the shift in O from the stored reference values.
|
