Binocular Vision System Of Space Robot For Non-cooperative Target

With the development of human space exploration, as an effective tool for on-orbit service, increasingly importance has been attached to space robot. It can not only reduce the risk of astronauts in orbit services, but also play an important role in satellites recovery. Aimed at free-floating space robot of visual servo system, this paper introduced algorithms such as camera geometric relationship, objects recognition of non-cooperative targets, stereo vision and so on. By establishing the free-floating space robot’s kinematics and dynamics model, it completed the PBVS system, details as follows:First of all, this paper deduced the camera’s internal and external matrix, then it accomplished camera calibration using images of a planar checkerboard. Based on Open GL, it then built a virtual modeling platform, which can be used into binocular vision and visual servoing.Secondly, in order to adapt to different non-cooperative targets, we put up with an object recognition algorithm based on the bag of words model. Through extracting SIFT features of the image, it then using the K-Means algorithm to get each "DNA of the images. By analysising of the "DNA" constitute of each image, we can tell the non-cooperative targets from each other.Then, we used stereo rectify to corret the images from camera distortion. After that, we can got disparity map by matching left image and right image. Using stereo reconstruction, the non-cooperative target’s position and pose can be obtained.Finally, this paper analysised the free-floating space robot modeling. By using PD and Feedforward control algorithm, we established the path planning model of free-floating space robot. Compared with Sim Mechanics simulation results, the correctness of visual servo system got be proved.