Development Of Agricultural Multi-spectral Low-altitude Remote Sensing Multi-rotor Unmanned Aircraft System And Its Application On Crop Growth Monitoring

The rapid and efficient acquisition and analysis of crop growth information is one of the key technical issues in the development of smart agriculture.By analyzing the plant height and other information of crops in different growth stages,it can provide real and effective data support for application scenarios such as crop growth monitoring,disease and pest assessment,and yield prediction,which is of extremely important value.UAV remote sensing platform is widely used in agricultural information collection due to its fast response,good mobility,and high efficiency.However,there are a lot of problems in currently commonly used agricultural remote sensing unmanned aircraft systems such as low system integration,poor applicability,and low operational accuracy.In response to the above issues,this study aimed to obtain accurate crop growth information quickly and efficiently,and conducted research in five aspects:agricultural UAV aircraft design,flight control system,aircraft control algorithms,remote sensing mission equipment and remote sensing data analysis method.An agricultural dedicated multi-spectral low altitude remote sensing multi-rotor unmanned aircraft system（UAS）was developed which could obtain crop canopy map information quickly and accurately in an agricultural operating environment.Taking oil-seed rape as the research object,an analytical method for analyzing crop growth information based on visible and near-infrared remote sensing images were constructed.The main research content and results of this article were as follows:（1）Aiming at the limited applicability of current unmanned aerial vehicle to agricultural remote sensing operating environments,a lightweight foldable agricultural multi-rotor unmanned aerial vehicle adapted to agricultural operating environments was systematically designed,providing a reliable aircraft platform for the development of agricultural multi-spectral low altitude remote sensing multi rotor unmanned aircraft systems.Based on the flight principle of multi-rotor UAV,a method for estimating and analyzing the endurance of multi-rotor UAV was established,which integrated the takeoff total weight,rotor number,and powertrain parameters.On account of this,a multi-rotor UAV power system design scheme was constructed according to the design requirements.Based on the selected dynamical system,a foldable octo-copter aircraft frame was designed,and the static analysis of the arm foldable components was performed.A three-dimensional visualized model of the equivalent stress and displacement of the arm foldable components was established.On this basis,weight reduction optimization design was carried out.Aiming at the design requirement of extending the endurance time,the impact of different capacities of drone batteries on the endurance time of UAV was verified through flight test.Besides,the optimized UAV was compared with mature UAVs on the market for endurance performance testing.The results were shown as below.（1）After weight reduction design,the weight of the folding part of the arm could be reduced by 47%,and the maximum deformation of the part was 0.0023 mm and the maximum equivalent stress was 5.93 MPa under extreme working conditions with an external load of 39.2 N.which met the safety design requirement.（2）There was a positive correlation between battery capacity and the endurance time of UAV within a certain range.However,when the battery capacity exceeded a certain range,its growth tended to be flat.Therefore,based on the verification results,a lithium polymer battery with a capacity of 16000mah was selected as the power battery for the low altitude remote sensing system of a multi-rotor UAV.（3）The average endurance time of the UAV studied in this article was 20.3minutes which showed that the endurance performance of this system is close to and to some extent superior to mature commercial products on the market.This indicated that the aircraft could meet the demand for precise acquisition of agricultural information.（2）Aiming at the problem of low integration of current agricultural remote sensing unmanned aircraft systems,a highly integrated agricultural low altitude remote sensing unmanned aerial vehicle flight control system was developed by building a wealth of communication interface resources with a flight control computer as the core,deeply integrating peripheral hardware devices such as sensors and power modules.Based on micro electro mechanical system（MEMS）technology,the software and hardware design scheme of flight control system was constructed.Aiming at the terrible agricultural working environment,two sets of mutually backed up sensor hardware were designed to improve the reliability of the flight control system.Aiming at the problems of multiple types of external equipment and complex requirements for UAV in agricultural operations,a dedicated power management module and multifunctional communication interfaces were designed to achieve in-depth integration of agricultural multi rotor aircraft with mission equipment,with flight control systems as the core.Aiming at the problem of large vibration of UAS in agricultural environments,a vibration damping device for flight control system was designed and its performance was verified.The results were shown as below（1）When the aircraft was flying horizontally at a height of 25 m and a speed of 2.5 m/s,the flight control system without vibration damping device was subjected to significant vibration interference,and the vibration intensity of its roll and pitch axes exceeds the safety range,with significant vibration in the heading axis direction.（2）The vibration intensity of flight control system equipped with vibration damping device has greatly weakened during the same flight process,with an average vibration intensity reduced by more than 94%compared to flight control system without vibration damping device,meeting safety requirement.（3）Based on the self-developed UAS described above,a UAS attitude and position controller with automatic disturbance rejection function was constructed,and an agricultural UAV control simulation system was developed.This system realized the tuning of control parameters for agricultural remote sensing UAV in a computer environment,suppressed the impact of external air flow on the flight attitude stability of UAVs,and improved the accuracy of agricultural remote sensing UAV system operations.Based on the data fusion technology from multi sensors,a UAV attitude calculation method was constructed,which lays foundation of a control algorithm for agricultural low altitude remote sensing multi rotor UAV was designed based on cascade control strategy.Aiming at the difficult problems of tedious and costly adjustment of control parameters for agricultural UAV,software in the loop simulation platform for flight control system was constructed to achieve the adjustment of control parameters for agricultural low altitude remote sensing multi-rotor UAV in a computer simulation environment.The results were shown as below（1）It was feasible to apply the UAV control parameters tuned based on the flight control system control simulation platform to actual aircraft flight,effectively reducing the aircraft attitude angular velocity deviation.（2）The average value of pitch angular velocity deviation decreased to 1.59°/s,and the average value of roll angular velocity deviation decreased to 3.3°/s The actual pitch angle and roll angle of the unmanned aerial vehicle based on the adjusted control parameters could effectively follow the expected values calculated by the flight control computer.The average angular deviations of the pitch angle was 0.65°while the roll angle was 0.64°,which showed that control effect was good.（3）When flying an unmanned aerial vehicle along an automatically set route,the average horizontal position deviation reached 0.28 m,and the vertical position deviation reached 0.07 m,resulting in high operational accuracy.（4）A narrow band multi-spectral remote sensing task equipment adapted to the folding agricultural multi-rotor UAV aircraft frame has been developed,and on this basis,an automatic path planning algorithm and task strategy suitable for agricultural working environments were constructed,improving the the professionalism and accuracy of agricultural remote sensing UAS,and achieving efficient autonomous remote sensing image acquisition of UAV aircraft in agricultural working environments.Aiming at the problem of missing photos caused by the independent operation of mission equipment and agricultural multi-rotor unmanned aerial vehicle,a communication port between mission equipment and flight control system was constructed based on pulse width modulation signals,achieving coordinated work between aircraft and mission equipment.Aiming at the problem that coordinated turning of aircraft could lead to excessive changes in the attitude of mission equipment,a fixed-point turning strategy for aircraft was designed and implemented.On this basis,a fixed-point trigger mode dominated by the flight control system was designed and implemented,and the mission strategy was optimized.The results were shown as below.（1）Compared with the commonly used coordinated turn mode for UAV,the fixed point turn mode could effectively stabilize the attitude of mission equipment.The average deviation between the actual value of the pitch angle and the expected value of the remote sensing image collected in the fixed-point turning mode was 0.96°,and the average deviation of the roll angle was 0.63°,which met the needs of agricultural information collection.（2）The UAV could fly according to the flight path automatically planned by the computer,and the image shooting position could correspond to the planned shooting position one by one.Compared to the timing trigger mode,the fixed-point trigger mode based on the coordinated operation process between the UAV and the mission equipment could effectively reduce the deviation between the actual position and the planned position of the remote sensing image acquisition.The average horizontal position deviation reached 0.77m,and the operation accuracy was high.（5）Based on the self-developed drone system mentioned above,a processing flow and method suitable for precise analysis of crop growth information were constructed,and target recognition of crops was achieved based on UAV remote sensing images.The feasibility of establishing inversion models for crop height,vegetation coverage,and leaf area index based on visible light near-infrared orthophoto remote sensing image data collected by agricultural UAV was verified.Based on geographic information registration technology,image matching and data fusion were performed between oil-seed rape multi-spectral remote sensing images and high-definition RGB images.The effect of remote sensing image data obtained at different heights on the retrieval model of rape plant height and the impact of the combination of texture features and vegetation index features on the construction of vegetation coverage and leaf area index inversion models were studied.The results were shown as below.（1）Target recognition algorithm based on Transformer backbone network could effectively identify wheat heads(AP₅₀ reached 88.3%)（2）It was feasible to apply remote sensing data fused with multi-spectral images and high-definition RGB images to construct a oil-seed rape plant height inversion model.There were significant differences between the remote sensing image data obtained at different flight altitudes and the retrieval model of rape plant height.The inversion model established based on remote sensing data collected at a flight altitude of 35 m was the best（R²=0.57）.（3）The vegetation index information and image texture information obtained by this system can be used to analyze the vegetation coverage and leaf area index of rapeseed.Only using a single vegetation index or texture information to model the leaf area index and vegetation coverage is not effective.The fusion of remote sensing data in various bands and then modeling and analysis of crop growth information can effectively improve the prediction accuracy of the model.After data fusion analysis,the coefficient of determination of rapeseed vegetation coverage model R² reached0.92,RMSE reached 0.04,the coefficient of determination of leaf area index model R² reached0.75,RMSE reached 0.43,which means the prediction effect of the model was good.