Research On The Key Techniques Of Mature Rice Main Attribute Parameters Inverting Based On LiDAR Data

The sustainable development of agriculture depends on accurate agricultural technology.The premise and key to achieve precision agriculture is the detection of field crop attribute information.In this paper,based on LiDAR technology,construct an airborne LiDAR crop information measurement system to meet the needs of large-scale mature rice property detection in the field.Crop point cloud data preprocessing algorithms such as K-D tree construction,outlier removal,ground point and crop point classification are studied,and the plant height is calculated based on the elevation difference between the classified ground point and crop point.At the same time,according to the characteristics of rice stem and leaf serious crossing and unsuitable ear division,a rice panicle layer point cloud data classification method based on point cloud intensity and elevation was proposed,and a rice density calculation method based on LiDAR data was studied.Inversion calculation of rice plant height,density,yield and biomass and other attribute parameters.The main research contents include:(1)Constructed an airborne LiDAR crop information measurement system suitable for the detection of attribute information of crops in a large area.The principle and composition of the LiDAR measurement system are analyzed,and the hardware of the laser scanner,GNSS system and INS system is determined.The VLP-16 laser scanner and the SPAN-IGM-A1 combination navigation product are used on the UAV M600,each working time can reach 16 minutes.Realize the collection of laser,GNSS and IMU data.When the flight height is 25 m,the flight speed is 5m/s,and the overlap rate of the flight belt is 40%.One flight can achieve 115 square kilometers of field crop point cloud data acquisition.(2)The algorithm of point cloud data preprocessing,the algorithm of estimation the rice plant height,density,yield and biomass based on LiDAR data were studied.Aiming at the characteristics of the scattered distribution of point cloud data and the presence of noisy points,the K-D tree structure is constructed to establish the spatial topological relationship of point cloud data.The methods based on Statistical Outlier Removal(SOR)and Radius Outlier Removal(ROR)are compared.The SOR effect is better when the point cloud removal rate is the same,and the quality of SOR outlier removal is better.When the radius r = 5 and the threshold m = 2 the SOR Works best.Through the grid division method,the rice plant height is calculated according to the difference in height between the crop point and the ground point in each grid area.Density calculation is based on clustering the point cloud data of the rice panicle layer that based on the combination of point cloud intensity and elevation,and calculating the rice density based on the number of point cloud clusters.Then,based on rice plant height and density,a yield and biomass estimation model is proposed.(3)Conducted different rice density distribution experiments in the field,and evaluated the accuracy of the rice plant height,density,yield and biomass estimation algorithm based on LiDAR data.The average root mean square error(RMSE)of the calculated plant height of rice and the field measured value is 6.771,the mean absolute percentage error(MAPE)of the calculated plant height of rice and the field measured value is 7.57%,and the accuracy is higher.In terms of density calculation,two clustering algorithms are compared.The RMSE of the calculated value of the super-voxel clustering algorithm and the field value is 9.968,the MAPE is 5.67%,the RMSE of the calculated value of the mean-shift clustering algorithm and the field value is 5.877,the MAPE is 3.37%,both algorithms are suitable for the rice density calculation in the field,and the accuracy of the mean-shift clustering algorithm is higher.The RMSE of the estimated yield and the field measured value is 0.014,and the MAPE is 3.83%.The RMSE of the estimated biomass and the field measured value is 0.069,and the MAPE is 5.06%.The accuracy of rice yield and biomass estimation based on this method is high,which can replace the results of manual measurement.