Experimental comparison and analysis of the performance of different image matching methods for different low-texture images

Image matching serves as a pivotal technical means for achieving efficient fusion of multi-source remote sensing imagery, thereby enhancing the overall data analysis capabilities. Among the commonly utilized feature-based image matching methods, SIFT, SURF and ORB algorithms have been widely used in the fields of image stitching, target detection and recognition for high-resolution remote sensing images. This paper delves into the performance of these three image matching methods for five sets of low-texture images such as indoor, underwater and so on. A comparative experimental analysis was conducted from the perspectives of image matching accuracy and image matching rate to evaluate the performance of these three feature matching algorithms on low-texture images. Based on the experimental results, we conclude that the ORB image matching method is not recommended for low-texture image matching scenarios that require high matching precision, and in engineering applications where speed is not a primary concern, the SIFT method outperforms both SURF and ORB algorithms by exhibiting stronger resistance to low-texture conditions.