Research On Visual Navigation Method Of Intelligent Combine Harvester

Accurate extraction of crop harvesting navigation path lines is becoming increasingly important for precision agriculture.The navigation path line can not only provide information for automatic driving system of the intelligent combine harvester,but also provide the real-time harvesting width for the yield of the harvester.Aiming at the problem that the navigation path detection of intelligent combine harvester is susceptible to interference and poor reliability,the complex working environment of the harvester is analyzed,and the visual detection model of the navigation line of intelligent combine harvester is established.A harvesting navigation line detection method based on aerial view navigation line visual detection area image segmentation is proposed,which solves the problem of obtaining the relative positioning parameters accurately and efficiently in the navigation system of the intelligent combine harvester.The navigation line detection algorithm based on the image segmentation of the harvested and unharvested areas is studied,and the field algorithm verification of the visual algorithm navigation line detection success rate and navigation parameter measurement accuracy is conducted in different regions and different environments.The main research contents are as follows:(1)Rice or wheat image information acquisition: According to the structural characteristics of the intelligent combine harvester and the characteristics of the harvesting environment,the type selection and installation position of the visual sensor are determined.The intelligent combine harvester and the visual coordinate system are established respectively,and the coordinate relations of different coordinate systems are matched according to the position relations of different coordinate systems.Then,the reasons and correction methods of image distortion in the camera imaging process are analyzed,and the image distortion is corrected using camera calibration parameters.Finally,the image is transformed by using the position of the static vanishing point of the image,and the image information acquisition target area is set in the image after the inverse perspective mapping(IPM).(2)Image segmentation of visual detection area of navigation line: The initial navigation line area is established by ultrasonic distance measurement.In the navigation line area established by ultrasonic measurement,the visual detection area of navigation line is set.Furthermore,the detection area of the navigation line is dynamically set according to the detection result of the navigation line of the adjacent frame image.Analyze image preprocessing algorithms such as inverse color transformation,histogram equalization,Gamma transformation to enhance and filter different target images.In the field test environment,the image segmentation algorithms such as Otsu and seed point region growth are used to segment the harvested and unharvested regions.(3)Visual navigation line detection algorithm of harvester: The application of random sampling consensus(RANSAC)algorithm and probabilistic Hough transform(PHT)algorithm in navigation line detection is analyzed.According to the image features of the navigation line visual detection area image segmentation,the deficiencies of the two navigation line detection algorithms in the navigation line detection are improved.Then,the detection effects of the algorithm before and after the improvement are compared.By using the image pyramid optical flow tracking algorithm,the corner points at the boundary of the harvested and unharvested regions are selected as the feature tracking points,and the visual navigation line of the harvester is dynamically tracked and detected.(4)Visual navigation test verification: According to the measurement characteristics of the visual navigation parameters of the intelligent combine harvester,the evaluation standards for measurement accuracy of angle and displacement deviation are determined.In different regions,light and locations of farmland,the visual detection algorithm of the harvesting navigation line of rice or wheat was verified by field experiments.Finally,the reasons for the failure of navigation line extraction and the error sources of navigation parameters measurement are analyzed.Experiments demonstrate that the visual navigation algorithm in the paper can meet the navigation path detection requirements of the harvester during the operation process.By improving the original detection algorithm of visual navigation lines,the problems of poor reliability and time-consuming in the navigation line detection process can be effectively avoided.