Research On Key Technology Of Autonomous Navigation For The Port AGV Based On Monocular Vision

The port container throughput of China has been ranked first in the world for the last ten years. However, it brings unprecedented pressure to the traditional mode of port transportation. Overseas ports have applied AGV System since 1990 s, which greatly improves the transportation efficiency of the port. Based on fixed path mode, guidance system applied in the port AGV at present is simple and reliable, but with poor flexible and intelligence.In recent years, with the development of image processing technology and computer technology, the vision-based navigation has attracted widespread concern. This paper builds a monocular vision system based on NI Starter KIT mobile robot to solve the problems of automomous positioning, path planning and obstacle avoidance of the port AGV, and it’s an exploratory research aimed at realizing the fully autonomous navigation of the port AGV. The main contents of this paper are as follows:(1) The design of positioning system based on monocular visual odometry. We use a single camera to acquire continuous images to accurate position estimation of AGV. Focused on improving the performace of SIFT feature points extraction and matching, we propose a monocular visual odometry which has high accuracy and a good performance of real-time,and then we can get the position of AGV in real-time by a single camera.(2) Global path planning for the port AGV. According to the structural environment of the port, we take into account the enery consumption and safety of the AGV to modifiy the A* algorithm, and then design a global path planning algorithm which is suitable for the port environment.(3) Local path planning for the port AGV. With the acquired image of road surface as samples, we propose an obstacle detection algorithm with adaptive image segmentation based on color feature of the road surface. We use monocular geometric positioning technology to determine the location of obstacle, and then select the best course angle based on the position of sub target and the obstacle position.(4) The design of hardware and navigation software system. We build a port AGV experimental platform and set up a local network by using the D-link-DIR617 wireless router, and then design both autonomous navigation and remote control software based on Lab VIEW and MATLAB.(5) We complete both of the remote and automatic navigation experiments of the AGV in the simulation of port environment. In the remote control mode, the image information of the road ahead of the AGV according to the IP camera is acquired, and then we use Lab VIEW to publish the shared variable to control the AGV. In the automatic navigation mode, the port AGV complete the autonomous localization, path planning and avoid the obstacles just by the image from the IP camera, and the results show that the autonomous navigation algorithm based on monocular vision system is robust.