Class-associative distortion-invariant pattern recognition using fringe-adjusted joint transform correlation.

摘要

Class-associative target detection is a recent concept in the area of pattern recognition. Alam and Rahman introduced a class-associative correlation filter for detecting a class of objects consisting of dissimilarities. However, if the targets are some what distorted due to variation in scale, rotation, or occlusion, this filter will fail in detection. In this research work, we applied the synthetic discriminant functions concept to enable the class-associative fringe-adjusted JTC to accommodate these distortions of the target. The feasibility of the proposed technique was tested via computer simulation using a database of stored images.