Research On Vision Navigation Technology Of Agricultural Robot In Vomplex Environment

With the development of computer technology and machine vision technology,visual navigation technology has gradually become the core control technology of intelligent self-propelled agricultural robot.The outdoor farmland environment has the characteristics of uneven illumination,vulnerable to extreme weather and uneven terrain,which interferes with the image acquisition and processing of the agricultural robot vision navigation system,resulting in inaccurate algorithm or inability to adapt to the current environment.This paper studies the farmland visual navigation algorithm in the outdoor complex environment,and puts forward a set of farmland headline detection and path extraction algorithm,which solves the problems of poor adaptability and low real-time performance of the traditional algorithm when the agricultural robot works under the conditions of uneven natural lighting and low visibility.Firstly,compared with structured farmland or greenhouse environment,unstructured farmland navigation algorithms have higher requirements for robustness and environmental adaptability.Some visual navigation algorithms applied to structured farmland are difficult to adapt to unstructured farmland environment.To solve this problem,this paper proposes an illumination compensation algorithm based on YCr Cb color space,which reduces the interference to the image when the illumination is too strong or too dark,and then uses the a priori knowledge of dark primary color to remove fog to obtain a clear image;Then,using OTSU binarization segmentation algorithm which reduces the traversal range,segmentation algorithm and morphological filtering combined with color tracking based on HSV color space,the binarization image with fine edge is obtained,which reduces the image segmentation time by about 40% and enhances the real-time performance of the algorithm.Secondly,in order to extract the walking path between farmland ridges,a feature point detection and correction algorithm based on vertical projection method is proposed in this paper.The characteristic points of ridge and ditch center in farmland are obtained,and a new straight-line fitting method is proposed by using random sampling and voting mechanism.This method can better extract the correct edge information and inter ridge feature points,and overcome the disadvantage of weak adaptability of traditional line detection algorithm in farmland path extraction;At the same time,using the idea of super-pixel,a farmland headland baseline detection algorithm is proposed,which can be used for headland position detection in various environments.Finally,the hardware platform of agricultural robot and the information processing and conversion process of agricultural machinery are studied.Through the conversion of image coordinates and visual coordinates,the calculation methods of robot pose,offset and offset angle are obtained,the visual navigation algorithm studied in this paper is simulated,and a set of simple software system is designed.