Research And Design Of Vision Localization System Of Aircraft In Space Cabin

The vision localization of mobile robots is one of the pillar techniques in the field of artificial intelligence and computer vision. It is a prerequisite for the robots’ autonomous motion and behavior decision-making. Without localization, robots don’t know where they are and where they want to go, and robots’ autonomous control is impossible.The study of the technology of visual position will help to improve intelligent and manipulative levels of robots.This paper studied and designed a monocular vision localization system based on a planar target, taking the aircraft in the space cabin as the research objective. The system can use the camera to capture actual scene images and realize the visual localization of robots by image processing and analysis technology. It can help the decision-making system of the aircraft control its motion, scientific experiments and tasks are achieved. The research of the visual technology in the design of aircrafts in the space cabin not only extends the application of machine vision technology, but also promotes the development of the space science technology in our country.This paper has carried out the following research work:(1) this paper reviewed and summarized the research status of the aircrafts in space cabin and the technology of visual localization, proposing a framework of the monocular vision localization system, and the system operation process are introduced. The framework built from the perspective of image processing and vision measurement, including the acquisition of images, camera calibration, target detection,3D pose computation, etc. After images had captured, the detection of targets in the images are carried out, and then the system, combining with the camera model,computed the position of target in ralative to the camera.Finally the system realized the aircraft’s vision localization.(2)The basis of visual measuring theory-the camera imaging model is studied, and the camera calibration method based on the 2D planar target is applicated to solve the internal and external parameters of the camera imaging model. The experimental results are given.(3)The target detection based on SIFT are studied and achieved, which includes the design of the target, the extraction of feature, image matching and RANSAC algorithm, etc. In order to improve the real-time performance of the target detection, the processing of the feature matching also is optimized, which reduces the redundancy of computation.(4) The P4P problem based on the rectangular distribution is studied, and the concrete calculation process is given. Finally, the 3D pose measurement of the objective is finished.The monocular vision localization system studied in this thesis can also be used in other situations where visual measurement is needed.