Adaptive dual-point Hough transform for object recognition

摘要

The dual-point generalized Hough transform is one of the most powerful tools for object recognition. Usually a single policy is used for all types of shapes, which may give rise to problems for objects with specific shapes. A shape dependent scheme can provide us a better R-table, and consequently reduce the number of point pairs for the accumulation process. In this paper, an adaptive approach is proposed to determine the policies for selecting the point pairs such that the resultant voting process is able to reduce spurious peaks and the point selection process is made more efficient.