基于三维激光点云的靶标叶面积密度计算方法

为向变量喷雾系统施药量的计算提供数据基础,提出了靶标喷施区域叶面积密度参数的计算方法。靶标三维点云数据由二维激光雷达传感器沿果树行直线运动间接获取。在假设各喷施子区域内叶片面积变化相对较小的条件下,基于Matlab曲线拟合工具箱cftool分析并验证了点云数与叶片数之间存在函数关系。曲线拟合结果表明,利用高斯函数、多项式函数与指数函数拟合点云数与叶片数,决定系数分别为0.925 7、0.931 0与0.936 4,指数函数拟合效果最好。相对误差分析结果表明,基于3种拟合函数,枝叶茂密区域相对误差最小为11.46%,枝叶中等茂密区域相对误差最小为11.05%,枝叶稀疏区域相对误差最小为35.50%。基于确定的点云数与叶片数间的函数方程,经系数变换后可计算出叶面积密度参数。

Calculation Method of Leaf Area Density Based on Three-dimensional Laser Point Cloud

Geometrical characteristic of the target is the basis to calculate pesticide dosage. In domestic and international research, volume was calculated usually as the control input of variable-rate spraying system. Compared with the volume parameter, leaf area density can avoid the errors caused by irregular canopy shape and lots of gaps in the canopy. Therefore, a method to extract the leaf area density based on the three-dimensional laser point cloud data was proposed. This method corresponded to the law that the branches and leaves were more, the leaf area was larger and the point clouds were more. A target detection system was set up, including a laptop, a light detection and ranging sensor, an incremental photoelectric encoder and an electronic slide. Three-dimensional laser point cloud of the target was obtained through the detection system moving along the line between the trees. According to the structure of the spraying system, the spraying area was partitioned into several spaces with the same scale, and the number of point clouds and leaves was counted respectively by computers and men in each sub-region. Based on the assumption which the leaf areas were almost the same in each spraying area, the function relationship between the number of point clouds and leaves was analyzed based on Matlab software cftool. The curve fitting results showed that the correlation coefficients between the leaf number and laser point number were 0.9257, 0.931 and 0.9364, respectively, with Gaussian function, polynomial function and exponential function, and the exponential function was the best. On the other hand, based on the above three functions, the minimum relative error of the thick area was 11.46%, the minimum relative error of the area between thick and sparse was 11.05%, and the minimum relative error of the sparse area was 35.5%. Then the leaf area density was calculated by the decided function and coefficient transformation. The calculation of the leaf area density from three-dimensional laser point provided a extra information for calculating the pesticide dosage. A statistic analysis on the real fruit trees and a optimizing for the target detection system were needed, more accurate function relationship between the laser point density and leaf area density would be calculated via the added correction coefficient.