| Hilly mountains region nearly covers 70% land area of China.Its agricultural production efficiency is quite different from that in plain areas.In hilly mountains region farmers mainly plant economic fruit trees.Restricted by terrain conditions,modern ground agricultural machines are difficult to use there.The mechanization level is much lower in hilly mountains region than national average mechanization level.Multi-rotor UAV with flexible movement has strong advantages on vertical take-off and landing,which are suitable for information collection and pesticide application in mountain orchards.But the precise control of UAV flight trajectory has become the bottleneck of efficient plant protection in mountain orchards.In this paper,according to the plant protection flight operation characteristics in mountain orchards,the UAV autonomous navigation flight control method based on the four-rotor UAV dynamic model and the multi-source information fusion was analyzed,.According to the distribution characteristics of fruit tree canopy in mountain apple orchard,a fast segmentation algorithm of fruit tree line based on machine vision was proposed,and a trend line extraction algorithm of fruit tree was explored.By combining the trend line extraction algorithm with GNSS navigation technology,an UAV horizontal track control method in mountain orchards based on GNSS-Vision fusion was proposed.According to the topography characteristics of mountain orchards,a three-dimensional point cloud extraction algorithm based on binocular image was adopted,and then,a distance measurement method based on binocular vision was explored.Aan UAV imitation flight control method based on binocular vision was also proposed.By a multi-source information fusion from horizontal track control and high imitation control,an UAV autonomous navigation method in mountain orchards was innovatively proposed.On the basis of the above theoretical research,an UAV autonomous navigation control system was designed,developed and verified through field test.The UAV autonomous navigation method in mountain orchards based on multi-source information fusion has important application value to improve the efficiency of plant protection in mountain orchards,promote the mechanization process of hilly mountain region,expand the scope of agricultural aviation of plant protection,and realize saving the cost and increasing the efficiency of the industry.The main research contents and conclusions of this paper are as follows:(1)The four-rotor UAV mathematical model of dynamics was studied,by assessing the characteristics of fruit tree planting in mountain orchards.Optimal flight path was analyzed,and the UAV autonomous navigation control method in mountain orchards was proposed according to the characteristics of flight trajectory.The navigation control was decomposed into a combination of two simple control methods: horizontal track control and height imitation control.The optimal UAV azimuth control and height control parameters were calculated by using PID control algorithm combined with dynamic model.Finally,a PID controller was adopted in azimuth control and a PD controller was adopted in high control to determine the azimuth and height respectively.(2)The UAV track control method of plant protection in mountain orchards based on GNSS-Vision fusion was researched.Based on the analysis of azimuth adjustment process in horizontal track control,the control system design was carried out.The work flow strategy was set as that the track was adjusted in operation line using vision navigation and the track was adjusted between the lines using the GNSS navigation.A RTK-GNSS device was built with two NEO-M8P-2 chips,and the host computer software was developed on the PC to upload and record the location information in real-time,meanwhile,the yaw information was obtained by calculating the real-time position and the preset position to realize azimuth adjustment.The fruit tree line was divided into three types color space of RGB,Lab and HSV,through the comprehensive evaluation of segmentation effect and processing time.A method of fruit tree line extraction based on RGB component linear combination in RGB space was selected,the trend line was obtained by two curve fitting of the segmented row,and then the yaw angle value was calculated for the azimuth adjustment.The horizontal track control of the UAV operation in mountain orchards was realized by combining two navigation modes.By verifing the accuracy of positioning and visual navigation,the results showed that the static positioning error of GNSS module was less than 0.26 m,the maximum error of dynamic testing was 0.82 m,and the average error was 0.53 m.The trajectory error of visual navigation was-27 ~ 48 cm and the average absolute error value was 23 cm respectively.(3)An UAV flight height imitation control method based on binocular vision was studied.Due to the poor accuracy of manual remote control and the non-fixed canopy height affected by the falling airflow disturbance,binocular vision technology was used for flight height control in view of the difficulty of maintaining the consistency of operation height.The binocular vision model was analyzed,the improved two-step calibration method based on OpenCV was designed and the camera was calibrated by using this method with a calibration error of 0.3489.To research the binocular vision stereoscopic matching and spatial positioning algorithms,the BM algorithm and SGBM algorithm were compared in OpenCV.The results showed that the execution efficiency of BM algorithm was up to 80ms/f,which was obviously better than that of the SGBM algorithm.Thus,the BM algorithm was selected for stereo matching in this paper.The UAV imitation flight control method was designed and implemented.The control accuracy of imitation flying control algorithm based on binocular vision was verified.The results showed that the binocular vision module had a good effect within 6m with a maximum relative error of 2.04% and an average relative error of 0.36% respectively.On another condition of 2.5m flying altitude,the average error of imitation flying control was 0.01 m and the maximum error was 0.15 m.(4)An UAV autonomous navigation control system was designed,which consisted of a flight platform and a ground control station.The size of the independently designed flying platform was 720×720×320(mm),with load capacity of up to 4kg.NAZA commercial flight control was chosen as internal control flight control.The autonomous navigation system was designed and developed.The GNSS navigation module,visual navigation module,copying flight control module and ground control station were developed respectively.The navigation system was also tested in orchards,and the results showed that the designed system could meet control requirements well.(5)An integrated test as well as the analysis of the autonomous navigation system were carried out.The system flight parameters of each module were analyzed and determined.The camera inclination parameter was determined to be 46 degrees.When the flight altitude was 2.0m,the field of view of the camera was 18.7m.The average processing speed of the extraction algorithm for testing the fruit tree row trend line was 6.53 fps.The control rate for visual navigation was determined to be 3 times per second.The results of the autonomous navigation experiments in mountain apple orchards showed that when the speed of the UAV was 2m/s,the altitude of the UAV was about 2m from the canopy of the fruit tree,the camera angle was 46o and the image navigation control rate was 3 times/second,the absolute error of track control was-47 ~ 42 cm,and the average absolute error was 19 cm. |
PDF Full Text Request