Model based robot vision

A robust methodology is presented for sensing and data processing to recognize industrial objects in a scene. The method is based upon an extension of the logical sensor concept. Each module of the sensing system is equipped with tests on its results to verify the correctness of the operation. If these tests are not successful, corrective action is taken. This is realized by adjustment of the module parameters or incorporation of another module in the processing. The modules and tests to recognize industrial objects in an image are described. The robustness of the method for noise and shading is demonstrated and compared with a conventional approach.<
>