Vision Navigation System Simulation And Walking Control System Design In Auto-walking Robot

Now, in China the picking work of vegetable and fruit in field is finished with handwork. In recent years, to solve the problems which are caused by the accelerating of population with aging and urbanization as the reducing of agricultural labor, increasing of agricultural production cost and decreasing of labor productivity, the automation of agricultural machinery must be improved. Another hand, the renewing and developing of agricultural machinery intelligence is promoted with the maturity of electronic technology, computer technology, image processing technology, artificial intelligence and the appearance of new production mode, so the research of machine vision and agricultural robot which is emphasized is an important direction in the exploring research of agricultural machinery intelligence and automation. The agricultural robot which is different from the industrial robot working in the certain environment mainly works in the natural scene, needs to face complicated and changing situation and has a lot of problems to solve. The non-structural field scene brings difficulties to agriculture picking robot. This experiment is aiming at the picking robot in the field to do the working environment recognition and moving control. According to the different function, the system is divided to two parts which are vision navigation and moving control.Many researches show that the vision navigation way that has advantages, such as, wide signal diction range and complete information acquisition is the mainly developing direction of robot navigation. Among the vision navigation ways, the navigation way which based on the local vision, installing the camera on the robot, is widely used. Therefore the navigation way of local vision is used in this experiment, and all the calculation devices and sensors load on the robot vehicle body. According to the judgment result of vision navigation, the moving control system chooses embedded Linux which is transplanted to the hardware platform of embedded operator as the operation system of platform and uses the design method of embedded system structure to construct embedded Linux crossed developing environment. The programs of image collection, preservation, display and motor driving in the embedded Linux system in terms of the characteristics of Linux driving program frame.The concrete research content of this paper is showed as follow:1) Image collection: The camera is loaded on the robot vehicle body, then, the direction mark image in lab environment and the furrow image in field are collected to be ready for image preprocessing; 2) Image preprocessing: The image collected can not meet the experiment requirement, so in the MATLAB environment, the collected image is processed with some image preprocessing, such as threshold separation, median filter, edge detection, expansion and corrosion to lay a foundation for judge the mark image direction and the furrow extension path;3) Experiment simulation: The feature extraction of mark image is extracted as the input of BP neural network training model. The direction of mark image and the furrow extension path are judged after the simulation experiment under MATLAB in PC to be the basis of robot moving;4) Hardware and bottom design of embedded system: The hardware part of control system is composed with camera, embedded control board and two-wheel driving vehicle. The embedded Linux crossed developing environment is constructed after connecting the hardware parts;5) Software design of embedded system: According to the characteristics and function of image processing, the programs of image collection, preservation, display and motor driving are compiled in terms of the program structure of Linux operation system;6) Whole debugging of experiment: According to the result of MATLAB experiment simulation and the embedded hardware and software system design, the whole experiment debugging which the auto-walking robot calls motor driving program module in terms of mark image direction information is done.All the experiment content above has done in this experiment. The original image which is collected by camera is processed with image preprocessing to obtain the image information which is needed by experiment under MATLAB in PC; the direction judging result is simulated in MATLAB experiment by constructing BP neural network training model; the furrow image collected in field is processed to get the clear furrow extension path as the basis of agricultural robot moving; the programs of image collection, preservation, display and motor driving are compiled in the embedded Linux crossed compiling environment. The auto-walking robot which real-time and intelligence are improved by control system can move freely in terms of the mark direction after the experiment debugging. The control system has some advantages, such as small hardware volume, modularization, easy extension, strong software real-time. The image processing algorithm task is added to Linux operation system and a lot of application developing interfaces are provided by hardware platform to be a stable basis for the further research of agricultural robot.