Researchers compare SIFT and ORB algorithms for matching satellite imagery using GPS-annotated test data and homography verification

Study evaluates two classical image matching techniques—SIFT and ORB—on satellite imagery for robotics, remote sensing, and geospatial applications

Both algorithms tested through a standardized pipeline: keypoint detection, descriptor extraction, matching, and RANSAC-based geometric verification with homography estimation

Matching quality measured using Inlier Ratio metric, which quantifies the fraction of correct correspondences in the estimated homography

Research uses a manually created dataset of GPS-annotated satellite image tiles with intentional overlaps to assess real-world performance

Analysis investigates how the number of extracted keypoints impacts the resulting Inlier Ratio for each algorithm