抄録
The image matching technology is important in robot vision. This technology is very useful in the case that images are transformed in parallel or are slightly rotated. However, the template matching can't be applied to the image with rotation, scale conversion on projection-distortion. In this paper, how to acquire the space position information of the target mark on the image is stated, taking the method with the reinforcement learning by which the right correspondence is obtained after the acquired image with rotation, scale conversion, or projection-distortion is projective back transformed.
