The application of image mosaic in information collecting for an amphibious spherical robot system

Abstract

Amphibious spherical robots that achieve the correct navigation need to collect surround information and clear perception. In addition to equip various sensors, the robot collecting information can also rely on the video images, which include rich texture and information of color. In this paper, we present a video image mosaic method to collect surrounding information of the amphibious spherical robot. We extract key frame from video image based on K-means clustering, and control the number of key frame by changing threshold. Meanwhile, pre-processing key frame and extracting the image key points based on SIFT (Scale-invariant Feature Transform). Then, use RANSAC (Random Sample Consensus) to eliminate the mismatch, and choose suitable value of scaling parameter of SIFT to improve match accuracy. Finally, stitching these key frames up, we got a complete image. Experimental results indicated that image mosaic not only conclude key points of video but also reduce fuzzy phenomenon, enhance the perception of direction. Compared with the direct splicing of video frames, the computation is greatly reduced and still retains important information.