Improved ORB Image Matching Algorithm Based on Local Information Entropy

Aiming at the problems of uneven and inaccurate extraction of feature points as well as low matching correctness of traditional ORB feature matching algorithm, an improved ORB image matching algorithm based on local information entropy is proposed. First, a Gaussian difference pyramid is established and meshed, and the local information entropy is used to filter the image meshes with rich texture information to improve the detection efficiency. Then, high-quality feature points are detected using the adaptive threshold FAST algorithm, and it is proposed to utilize the local information entropy occupancy relation to distribute the feature points uniformly to reduce the false matching rate. Finally, the BEBLID binary descriptor is combined with the improved ORB feature detection algorithm to further enhance the correct matching rate. The experimental results in the Oxford standard dataset show that the feature point uniformity of the proposed improved algorithm is more than 21% higher than that of the ORB algorithm, and the operation speed and matching effect are significantly better than those of other similar improved algorithms.