Real-time object matching
Aiming Huang, Zheng Gao, Bin Dai, Luo Li · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1998
In this paper, we present a high quality algorithm for object matching. Our goal is to find the position of a matched image(a model) within a reference image (it is larger than the model, and it can be decomposed into many samples). We just focus our attention on the shape to speed up matching, which based on sets of edge maps associated with the objects of interest. The order-statistic filters is applied to detect edges of the reference and the model. It can extract more uninterrupted edges both of the strong and the weak .Then the modified Hausdorff Distance is used to decide the similarity between two objects, and it costs less in computing the measure of similarity between two objects. At last we propose a pyramid structure to analysis the objects in several resolutions. It is a coarse-to-fine registration and it can make the number of computations tractable. The algorithm was applied to several real objects which were segmented from satellite maps and plane maps. We have shown that it has some desirable properties to satisfy our objective.