Research On Calibration And Super-resolution Of Vision Preprocessing Based On Space Soft Robot

With the development of space technology,the tasks of robots applied in space operations have become more and more complicated and diverse.There are many tasks such as flexible crawling of space non-cooperation targets,narrow space exploration,pipeline repair,complex 3D terrain crossing,etc that traditional structural manipulators are difficult to complete.Thus,the space soft arm is a new type of space robot that comes into being for complex space tasks.As the most important input source of information,the space soft arm puts more requirements on the visual system,such as more accurate positioning,more stable outcome,etc.For these needs,preprocessing algorithms of the vision system become especially important.The aim of this paper is to select from various types of image preprocessing algorithms and make related improvements on the basis of in-depth discussions about the soft arm vision system,which can provide an effective solution to the preprocessing process.The research of this paper is mainly divided into two parts: the first part mainly introduces the software and hardware system of the space soft arm and its vision system.The second part consists of the image preprocessing algorithms of the space soft arm vision system and its further research.The main work completed in this article is as follows:1)The introduction of the spatial variable geometry truss structure(VGT)and the space soft robot using shape memory alloy(SMAs)for active controller;The analysis of forward kinematics based on its SMAs characteristics.2)The design of the visual system and its image processing algorithm flow;The analysis of vision system preprocessing algorithms based on the soft robot's tasks and the improvement on each algorithm.3)The design of the camera calibration method as to the spatial soft robot vision system;The proposal of a new chessboard calibration method based on low rank texture and directional total variation(DTV).4)The design of an image super-resolution method as to the spatial soft robot vision system;The proposal of a new image super-resolution method based on compound regularization.Massive reliable experiments proves that the chessboard calibration method proposed in this paper retains the simple mode of operation and significantly improves the accuracy of camera calibration algorithm.In addition,the image super-resolution algorithm proposed in this paper has a good recovery effect for low resolution images in highly structured artificial scenes.