Overlapping object recognition: a paradigm for multiple sensor fusion
Intaek Kim, George Vachtsevanos · IEEE Robotics & Automation Magazine · 1998
Recognizing and identifying overlapping or occluded objects is a problem typically encountered in a manufacturing setting. The resultant image distortion tends to limit the applicability of current recognition systems in that case. The proposed recognition scheme involves the utility of an appropriate suite of complementary sensors and is based upon a systematic methodology that addresses the modeling problem through a polygonal approximation and the matching task between the sensor data and stored templates through a construction, called the intervertex matrix. An example is included to illustrate the simplicity and flexibility of the proposed approach.