Automatic image matching with point detection algorithms

The concept of image matching is one of the most important common topics in areas like Digital Photogrammetry, Remote Sensing, Image Processing and Computer Vision. Especially in Photogrammetry and Remote Sensing, image matching process must be carried out in order to determine interested point positions and to achieve the coordinates in image space with stereo evaluation. To match images, feature points must be selected on stereo images and the conjugate points must be identified in the best possible way. In this study, SURF (Speeded Up Robust Feature Extraction), FAST (Features from Accelerated Segment Test) and HARRIS (Harris & Stephens Detector) algorithms, three algorithms for image matching and point extraction, have been tested and compared against each other. Furthermore, the best matched common feature points have been matched automatically, and using the RANSAC (Random Sample Consensus) algorithm, the incorrect matches have been removed automatically. The resulting performances of SURF, FAST, and HARRIS algorithms, the three point extraction algorithms widely used in recent years, on Photogrammetry and Remote Sensing images have been presented. The FAST algorithm succeeded especially in catching the corner points and the SURF algorithm produced more feature points to be matched with a high precision. The algorithms used in this study, are evaluated to be very useful in the automatization of the basic tasks in image matching and 3D coordinate acquisition in Photogrammetry and Remote Sensing image analysis.