基于LIDAR技术的喷雾量三维空间分布测试方法

为解决喷雾量分布测试中耗时长、工序繁琐、无法进行实时动态三维空间分布测量的问题，该研究开发了一种基于激光雷达探测技术的喷雾量三维空间分布的测试方法。针对植保作业过程中常用的空心圆锥雾、防飘空心圆锥雾、扇形雾和防飘扇形雾4类共7种喷头，采用喷雾量实测方法对距离喷头50 cm处雾流区截面的雾量分布进行测试；利用十六线激光雷达对雾流区进行三维探测，实时获取喷雾量点云数据信息，通过数据包解析、仿射矩阵空间转换、坐标系解算获取点云坐标及密度，并利用神经网络将喷雾量实测结果与激光雷达测试结果进行拟合。结果显示，7种喷头训练集拟合相关系数r≥0.995，验证集r≥0.935，测试集r≥0.877，扇形雾喷头总体拟合相关系数r≥0.990，证明激光雷达探测是一种可行且准确的喷雾量分布测试方法；进一步对各喷头喷雾量点云数据进行分层网格化计算得到雾流区三维空间雾滴分布特征，结果表明3种圆锥雾喷头空心段长度大小依次为ITR、TR和HCI喷头，IDK喷头等距离喷雾截面积均大于LU喷头。该方法可准确地完成三维空间喷雾量化分析，同时也可为喷雾设备雾化质量检测、室内和田间雾滴飘移测量、植保机械田间快速调校及作业质量在线监测提供一种新思路。

Method for measuring the 3D spatial distribution of spray volume based on LIDAR

Abstract: Spray volume distribution in the three-dimensional (3D) space of nozzles is an essential interfering factor on spray drift and deposition of pesticide application, particularly on the atomization quality. Uniform distribution of spray can contribute to an obvious enhancement of pesticide efficacy, while reducing overuse and serious environmental contamination. However, the accurate measurement is still lacking in the real-time dynamic 3D distribution of spray volume, due mainly to long time consumption, and cumbersome procedure at present. In this study, a novel measurement for 3D spray volume distribution was developed using light detection and ranging (LIDAR) technology. Seven types of nozzles were tested, including the commonly-used nozzle of hollow cone, anti-drift hollow cone, flat fan, and anti-drift flat fan (HCI4002, TR8002, ITR8002, LU9002, IDK9002, LU12002, and IDK12002) in plant protection. The spray area of the nozzle was scanned using a 16-line laser LIDAR with the laser (Class 1) wavelength of 905 nm and the scanning range was -15°-15°. Specifically, the angular speed of horizontal rotation was 5 Hz, and the emission frequency was 320 Hz. The scanning lasted for 60 s, and all nozzles were tested with 3 replicates. The point cloud data was transferred to the laptop in form of packets in real time. MATLAB 2019b software was used to run the affine matrix and coordinate system transformation after data packet analysis for the droplet coordinates and spatial density. Meanwhile, the real value of spray volume distribution was measured in the spray section of 50 cm below the nozzle. Polyethylene (PE) centrifugal tubes with a volume of 50ml were arranged in a matrix to collect the droplets. Four kinds of fan nozzles were tested by a 5×15 collector matrix, and three kinds of hollow cone nozzles were tested by a 9×9 collector matrix. All nozzles were measured three times, and all tests lasted for 3 min, in order to collect enough droplets for a small weighing error. A neural network with 1 hidden layer (100 hidden neurons) and 1 output layer was used to fit the relationship between the traditional measurement and LIDAR scanning. The ratio between training, validation, and testing set was 70:15:15. The results showed that a high fitting precision was achieved in all seven kinds of nozzles for the correlation coefficient in the training set r≥0.995, validation set r≥0.935, testing set r≥0.877, and the correlation coefficient r≥0.990 for the flat fan nozzles. It proves that the LIDAR scanning can accurately and quantitatively analyze the spray volume distribution. The 3D spatial distribution of spray volume for all 7 nozzles was obtained after the spray area was layered and meshed, then to calculate the droplet density in each grid. A faster and easier procedure was made for the real-time 3D spray volume distribution, compared with the conventional one. The LIDAR technique can also be expected to provide an alternative way for atomization quality detection of sprayers, indoor and field test of spray drift, particularly on a rapid adjustment and online monitoring of operation quality in plant protection machinery in the field.