Research On Automatic Navigation Technology Of Unmanned Agricultural Machine Based On Image

Automated navigation of agricultural vehicles is an important part of precision agriculture,and visual navigation technology has the characteristics of low cost,high accuracy,and strong environmental adaptability.The agricultural machinery visual navigation combined with the two has considerable application prospects and development potentials.However,due to the impact of complex agricultural environments,agricultural machinery visual navigation faces urgent problems such as how to effectively extract paths,improve navigation accuracy,and accurately identify field obstacles.Therefore,this article focuses on the key technologies in agricultural machinery visual navigation,and conducts the following research step by step:(1)Camera calibration and distortion correction technology,establish a monocular vision imaging model,use Zhang Zhengyou's calibration method and coordinate system conversion to calculate the internal and external parameters and distortion coefficients of the camera,and achieve distortion correction by interpolation;(2)Target path extraction technology,comparing the effects of multiple graying methods,analyzing the limitations of EXG method combined with OTSU method for binarization,and presenting improved measures for repeated binarization based on EXG method and HSI color space Image segmentation is completed.For noise holes and non-target paths,morphological processing and chain code tracking are used for filtering;(3)Navigation point extraction technology.by analyzing the distribution characteristics of the edge points of the field path,an improved genetic algorithm based on the ring model is proposed,which can adaptively adjust the noise intensity.The accuracy of the navigation point extraction is more than 3 times higher than the traditional coordinate midpoint method;(4)Obstacle recognition technology,an improved HOG + SVM obstacle recognition method based on circularity is proposed,and the classification accuracy reaches 96%.According to the judgment process of rough positioning,re-identification,and later precise positioning,it can be used to accurately identify specific obstacles in stones in actual scenes,providing a theoretical basis for vehicle avoidance;(5)Design and build a software and hardware platform for simulating agricultural machinery trolleys,use DSP as the control core,and implement visual navigation functions in an embedded environment.The experimental results show that the simulated agricultural machine implements the visual navigation function on the path between forests.The maximum deviation from the center of the road does not exceed 4cm,and the root mean square error does not exceed 3cm.The accuracy meets the requirements of agricultural machinery navigation,and realizes target obstacle recognition and positioning in various scenarios.The feasibility of the improved algorithm in this paper is verified,and it provides reference and theoretical support for the research of image-based unmanned agricultural machinery automatic navigation technology.