Recognizing occluded flat 3‐D objects with internal motions based on the tracing of affine invariants
Jim Z.C. Lai · Journal of the Chinese Institute of Engineers · 1994
In this paper, a method for recognizing occluded 3‐D objects with internal parameters is presented. This method is simple and robust. In contrast to the conventional method, which uses an exponential model of the number of feature points to identify images, the present approach achieves image recognition via a simple linear function of the number of corner points. Therefore, the computations required to carry it out are much less complex. Simple heuristics are used to speed up matching between a candidate model and an image. The conventional method assumes fixed thresholds for all objects in the recognition process. It is shown that we cannot distinguish affinedly similar objects based on the affine invariants, thus the thresholds should be functions of the models. Here, a method of determining the threshold Ω i , of model i is presented. Finally, all the concepts presented in this paper are illustrated through several examples.