Feature Matching For Vision System In Forestry Robot

Forestry robot is a more efficient solution to solve the problem of labor shortage and low efficiency instead of human, conducting forest works. Feature matching is the key issue for robot binocular vision system, realizing forestry robotic intelligence.In this thesis, the main research achievements are as follows:1. This thesis presented a new point matching method based on color invariants in rgb orthogonal color space, using SURF algorithm. The experimental results show that the proposed matching method has a strong performance with evaluation criteria:more than 15% extraction point number,27% lower false matching rate, higher matching score, compared with SIFT, SURF algorithm in grey level and CSEFT.2. This thesis presented a new line matching method based on color invariants and the descriptor we presented similar to color histogram in the rgb orthogonal color space, using Hough transform. The method is innovative by applying color information to the Hough transform extraction and the line matching parts, for which there is almost no relevant literature or research. The experimental results, using the ALOI database and images captured on the Beijing Forestry University campus, show that our line matching method performs stronger than traditional methods with 20% lower false matching rate, higher matching score.3. A novel region matching algorithm based on the color invariants and color invariant moments in the rgb orthogonal color space was proposed, using maximally stable extremal region (MSER) and the nearest neighbour method to match color images. It can take advantage of both the color and geometric properties of the images, solve the problem of illumination influences, simultaneously the large computational time compared with the prior methods:computing every channel. The experimental results show that the proposed algorithm has a better performance than that of prior art with more than 30% extraction region number,35% lower false matching rate and approximate 50% matching score.