Image-matching framework based on region partitioning for target image location

The target-location problems of observation and combat-integrated UAVs utilized in battles makes image matching challenging and of vital significance. This paper presents a framework of image matching based on region partitioning for target-image location, working on complex simulated aerial images consisting of, for example, scale-changing, rotation-changing, blurred, and occlusion images. Originally, an image-evaluation approach based on a weighted-orientation histogram was proposed to judge whether the image is an image with good texture or a textureless image. Two approaches based on layered architecture are employed for images with good texture and textureless images. In these two approaches, an improved SIFT image-matching algorithm incorporating detected Harris corners into the keypoint set is suggested, and Bhattacharyya distance based on an orientation histogram was employed to select the best result among different region pairs. Experiment results illustrated that the image-matching approach based on image segmentation has a much higher rate of 42.04 when compared to the traditional approach.