2D template matching for automated inspection using generalized hough transform

In this paper, we modify GHT to reduce its memory requirement and computational cost and improve its performance under rigid motion. In the proposed algorithm, orientation and displacement are determined separately using a multi-stage scheme, and the displacement accumulator-which occupies most of the memory-is downsampled without degrading detection resolution. Moreover, the template pose becomes more reliable by using an adaptive weight scheme when incrementing. Experimental results show that the proposed scheme has advantages in both computational complexity and memory requirement over conventional GHT, and pose determination is more stable.