Computer Vision Matcher for Extracting and Detection Objects

Abstract

The research examines the issue of feature matching and object recognition in two images with the use of brute-force matches. The suggested framework utilized several concurrent algorithms, including BRISK (Binary Robust Invariant Scalable key points), SURF (Speeded-Up Robust Features), ORB (Oriented FAST and Rotated BRIEF), and SIFT (Scale Invariant Feature Transform), for descriptor extraction and feature detection. K-Nearest Neighbors (KNN) algorithm in conjunction with the brute-force approach allows for feature matching. The robust Random Sample Consensus (RANSAC) approach calculates the transformation between two successive images using the found matches. As a result, the RANSAC approach is used to enhance the removal of outliers. With the use of the OpenCV library, the suggested technique was developed and put into practice. Analyses of the speed at which it executes commands and the precision of the feature matching serve as proof of the system's quality and efficacy.