Research on Feature Point Extraction Algorithm for Drone Images

Feature points are extensively utilized in image processing and play an important role in image matching and 3D reconstruction. Many algorithms and detection operators have been proposed to achieve the extraction of the image feature points. We choose MSER, SIFT, SURF, ORB, BRISK, and AKAZE as the critical research objects to compare the application effect of algorithms. On the basis of a detailed analysis of the execution principles and processes of the six algorithms selected, we selected a drone image containing multiple types of ground objects as the experimental object, and extracted the feature points using the six algorithms respectively, and conducted a comparative analysis of their execution efficiency and the number of extracted feature points. The experimental findings indicate that SURF and ORB have high execution efficiency while ensuring the extraction of an adequate quantity of feature points. SIFT, BRISK, and AKAZE strike a balance between computational efficiency and the quantity of retrieved feature points. The computational complexity of the MSER algorithm is considerable, and the quantity of retrieved feature points is little.