Research Of Greenhouse Spray Robot’s Visual Navigation And Positioning System Based On Kinect

It is of great significance to study the application of greenhouse spraying robot. Greenhouse spray robot can save labor, avoid worker excessively exposing in chemical pesticides. Navigation and positioning technology based on computer vision play a great role for spraying robot in greenhouse walking independently and completing spraying task.In this paper, according to the principle and function of the Kinect sensor, I present constitute and research scheme of spraying robot vision navigation and positioning system. At a target-known greenhouse, a robot starts from a starting position, then self-locate using monocular visual odometry, and adjusts the walking direction according to the location; when near the vegetation area, robot can detect boundary between road and vegetation area through vision, and robot walks and sprays along the boundary; in the process toward target, using depth image information robot recognizes obstacles, can autonomously avoid obstacles. Accordingly, we focus on the rearch of monocular visual odometry, including research and improvement of features detection and matching algorithm, combined with the characteristics of SIFT and FAST, proposed a real-time algorithm with high matching accuracy; and the visual odometry positioning algorithm; Then this paper proposes road boundary navigation based on road recognition and boundary extraction, and obstacle recognition based on infrared camera, and puts forward the solution of behaviors fusion, which are target navigation, road boundary navigation and obstacle avoidance; the final two parts of experiments show that, the algorithm can guarantee the accuracy robustness and real-time performance of monocular visual odometry and the robot can safely complete the task of road boundary navigation and obstacle avoidance.