| Precision agriculture is to achieve maximum economic benefits through fine adjustment of crop management measures.The premise and key of precision operation is accurate detection of crop geometric characteristics.Based on the technology of Light Detection And Ranging(LiDAR),this paper constructs appropriate LiDAR systems respectively for the detection needs of crops in different greenhouse and field environments.When obtaining the height of individual plant,soybean crops planted on field ridges have the characteristics of large difference in ground flatness and serious crossing of branches and leaves.The data processing method of extracting point cloud of individual plant to calculate geometric characteristics was studied.At the same time,because the laser cannot collect the root of a single plant,the estimation method of the measuring point of the root of a single plant is studied.The method realizes accurate segmentation of single plant crops and extraction of geometric parameters such as height,volume and the like.The main research contents include:(1)A 3D LiDAR detection system suitable for efficient acquisition of field crop morphology is constructed.In view of the characteristics of large crop area,complex terrain and limited vehicle-mounted detection path when accurately and efficiently acquiring geometric features of field crops,a lightweight airborne LiDAR detection system scheme is proposed.The precision of VLP-16 laser scanner is 3cm,the data update frequency of SPAN-IGM-A1 integrated navigation measurement unit is 200 Hz,the frequency of differential GNSS(GPS+BDS)is 20 Hz,and the application of IE(Inertial Explorer)pose calculation software is exactly reflected.The maximum moving speed of the UAV platform is 18m/s,and the laser scanner can acquire 300,000 data points per second.After system integration,the total mass of the carrying part is about 3.6kg,which meets the requirement of airborne detection less than 6kg.Through the installation error calibration test,the installation attitude angle error of multi-source sensor is corrected,and the accuracy verification test is carried out.The precision verification test results show that the error value of the system in the system forward direction dimension of 1000 cm is 3.1cm;The error value of the dimension 400 cm perpendicular to the crop row direction is 6.2cm;In the direction perpendicular to the ground,the error value of angle 75 is 0.6,which can realize accurate detection of crop geometric characteristics.In order to give consideration to on-board detection,a universal mounting bracket and a wheeled remote control mobile platform for mobile measuring end are designed.(2)extracting individual soybean point clouds and calculating plant parameters.According to the characteristics of large difference in ground flatness of planting patterns on field ridges,a calculation method of extracting single plant crop point cloud before calculating crop geometric characteristics is proposed.Firstly,crop and ground are segmented,and a segmentation method based on local features of point cloud neighborhood is proposed.Secondly,aiming at the problem of branches and leaves crossing among crop individuals and difficult segmentation,the mean shift algorithm is used to segment individual plants with reference to plant planting characteristics.In order to estimate the measuring point at the bottom of the plant,the clustering center of the individual plant is projected onto the surface fitted by the ground point cloud.Alpha-shape algorithm is used to reconstruct crops in three dimensions so as to facilitate volume statistics of individual crops.The results show that the method based on local features of point cloud neighborhood can achieve the ground and crop segmentation effect with AUC value of 0.9944.When the bandwidth of mean shift algorithm is H=20cm,the optimal single plant extraction effect is achieved.The maximum relative error of crop height was reduced from 17.66% to-2.87%,and the maximum error of extraction was 7cm,which realized effective extraction of geometric parameters of individual crop.(3)A pendulum-type two-dimensional LiDAR detection system is constructed.In view of the lack of GPS signals in the greenhouse and the high cost of applying three-dimensional LiDAR,this paper constructs a pendulum-type two-dimensional LiDAR detection system according to the structural environment characteristics of the greenhouse and the detection requirements of variable sprayer.The system integrates steering gear,two-dimensional laser rangefinder UTM-30LX-EW,Attitude Sensor,odometer and other main hardware.Data fusion and software system development have been carried out,and the detection accuracy of the two-dimensional LiDAR detection system has been tested and verified.The results show that the maximum relative error of the system detection accuracy is-12.9%,and the maximum absolute error is 10 cm when the size is 77.7cm,which can be used for target application in greenhouse. |
PDF Full Text Request