树高是森林资源调查的重要参数,由于传统人工估算树冠树高方法存在费时、费力等缺点,因此,无人机激光雷达技术被广泛应用.首先,获取数据并分析,采用坡度滤波算法和布料模拟滤波算法提取地面点的精度情况,其相关系数R²分别为0.9927、0.9922,坡度滤波算法提取地面点准确度略高;然后通过自然领域法插值生成数字高程模型(Digital Elevation Model,DEM),再用归一化数字表面模型(Nor-malized Digital Surface Model,nDSM)提取冠层高度模型(Canopy Height Model,CHM),采用局部最大值算法和分水岭算法提取树高;最后,用实测结果进行精度验证.结果表明:树顶点识别与实测树顶点结果召回率96%、准确率100%、F测度为98%,说明树顶点识别精度较高;树高提取均方根误差达到0.33、最大绝对误差为1.87、最小绝对误差为0.01,说明轻小型无人机遥感技术可快速高效地测量单木树高,且测量精度满足林业调查基本需求,为城市林业资源调查提供技术支持,实现高效化和低成本化.
Analysis of Tree Height Extraction Based on UAV LiDAR Data
Tree height is an important parameter in forest resource surveys.There are drawbacks such as time-consuming and laborious processes of traditional manual methods for estimating tree crown height,unmanned aerial vehicle(UAV)LiDAR technology has been widely used.Firstly,obtain data and analyze the accuracy of extracting ground points using slope filtering algorithm and fabric simulation filtering algorithm.The correlation coefficients R ² are 0.9927 and 0.9922,respectively,indicating that the slope filtering algorithm has slightly higher accuracy in extracting ground points;Then,the Digital Elevation Model(DEM)is generated through natural domain interpolation,and the Canopy Height Model(CHM)is extracted from the Normalized Digital Surface Model(nDSM).The tree height is extracted using local maximum and watershed algorithms;Finally,validate the accuracy with the measured results.The results show that the recall rate of tree vertex recognition is 96%,the accuracy rate is 100%,and the F-measure is 98%compared to the measured tree vertex results,indicating a high accuracy of tree vertex recognition;The root mean square error of tree height extraction reaches 0.33,the maximum absolute error is 1.87,and the minimum absolute error is 0.01,indicating that light and small unmanned aerial vehicle remote sensing technology can quickly and efficiently measure the height of individual trees,and the measurement accuracy meets the basic needs of forestry investigation,providing technical support for urban forestry resource investigation and achieving high efficiency and low cost.
UAV LiDARlocal maximum valuewatershed methodtree height extraction
万方数据
维普
