Dynamic image stitching for moving object

Abstract

Environment perception is a key technology for robot control and moving. Image information has the advantage of low-cost to perceive the indoor environment. Stitching multiple images is helpful to realize the full range of global view of indoor environment. Our research devotes to showing a large field of view by dynamic image stitching when there is a moving object in the environment. The images of two ceiling cameras were stitched by using methods of image stitching. A core process of image mosaic is adjusting intensity and seeking optimal seam line. The optimal seam line was computed to avoid the moving object, which solved the abnormal stitching caused by parallax influence. Our experiment was divided into two stages: preparation and test. Initial setup got the stitching parameters and built background. This reduced time for the test stage. We designed an experiment of moving robot real-time location and control. We used this experiment to test our methods. The experiment results show that we can well achieve dynamic image stitching.