ORB Feature Extraction and Matching Algorithm Based on Sampling Probability Optimization

In response to the issues of missed detection, false detection, and matching clustering in the traditional ORB algorithm's feature point extraction, an improved ORB feature extraction and matching algorithm based on sampling probability optimization is proposed. Firstly, during the FAST feature point extraction process, a probability distribution model is constructed based on the distance between target pixels and matching pixels, such that matching pixels closer to the target pixels have a higher sampling probability. Secondly, in the feature matching stage, a sampling-based descriptor construction model and descriptor vector optimization strategy are introduced to enhance the feature correlation between the matching pixels and target keypoints, while improving the discriminative ability between different matching pixels with the same sampling probability. Experimental results show that the improved algorithm significantly enhances the feature point extraction accuracy and feature point matching efficiency compared to traditional methods, while also achieving a more even distribution of matching points.