Research Of Automatic Navigation System Based On Embedded Vision

The automatic navigation of agricultural vehicle is a key technology to realize the modern intelligent agriculture, which can lighten the intensity of labor, reduce the cost of production and improve the agriculture working accuracy. Our country is a large agricultural nation, with few farmlands and large population, the development of agricultural relatively lags behind, but also the poor level of agriculture machinery automation. in consequence, the research on automatic navigation of agricultural vehicle has great practical significance.This subject following the development of technology, combining embedded technology with image processing technology, researches the intelligent navigation system based on embedded visual. Focusing on the access of visual navigation lateral control parameters, this paper study the principle of vehicle positioning based on machine vision, the algorithm of navigation path extraction and the transplantation of algorithm to embedded hardware platform.Firstly, it is introduced the overseas and domestic research status about the technology of agricultural vehicle navigation, relative positioning technology and designing scheme of the system, Introduced the appropriate embedded hardware platform and image sensor according to the requirements of visual navigation for the research, briefly described the visual image processing library used in the research.Then, it discusses the principle and method of visual navigation agricultural vehicles, analyzes the relations between camera inner parameters and the gesture of vehicle, acquires the calculating method of lateral control parameters including lateral displacement and angular deviation,in the process of navigation, obtains the value of inner parameters and distortion parameter of the camera used in the searching.Take tea garden in natural environment as an example to illustrate the process of algorithm in detail, the algorithm includes position judgment, navigation path extraction in the crop line and navigation line recognition at the end of crop line. Through the experiment, it shows that the accuracy and real-time capability of navigation path extraction algorithm can satisfy the practical application, and the position judgment method is viable in some degree.It achieves the building of navigation system development environment on embedded platform, including the transplantation of Boot loader and the Linux kernel,the production of root file system, and the cross-compiler environment building,as well as the transplantation of Open CV. The navigation line algorithm was successfully transplanted into the embedded platform, and the feedback control data was sent out via a serial port.At last, combined the research situation, summarized the existing urgent problems in the research on automatic navigation of agricultural vehicle, and prospected the research hotpots, to provide theoretical basis and experimental reference for further research.