High density point clouds from unmanned airborne laser scanning (UALS) systems have great potential for small area forest inventories. We propose an UALS-based tree level inventory method that takes advantage of both the segmented crowns and segmented trunks: hybrid tree detection (HTD). The method is tested at twenty 30 m × 30 m validation plots of varying maturity, tree species distribution and stocking density. With the traditional individual tree crown delineation (ITC) approach, tree attributes are only predicted for crown segmented trees. Here, we assume that a segmented crown can contain more than one tree. Our idea is to identify segmented crowns that contain more than one segmented trunk. One of the trunks is linked to a segmented crown (upper most tree) and the remainders are treated as understory trunks. Heights of the crown segmented trees and the diameters of the understory trunks are used as predictor variables in nonlinear mixed-effects models of tree volume. The %RMSE and %MD values of volume predictions at the 30 m × 30 m validation plot level were 22.2%, −13.3% and 18.8%, −8.3% for ITC and HTD, respectively. We conclude that the proposed HTD approach improves the accuracy of ITC in managed boreal forests when using UALS data.

来自无人机载激光扫描（UALS）系统的高密度点云在小面积森林清查中具有巨大潜力。我们提出了一种基于UALS的树级清点方法，该方法可以同时利用分段的树冠和分段的树干：混合树检测（HTD）。该方法在20个30 m×30 m验证图上进行了测试，这些验证图具有不同的成熟度，树种分布和放养密度。使用传统的单个树冠轮廓描绘（ITC）方法，仅对树冠分割树进行树属性预测。在这里，我们假设一个分段的树冠可以包含不止一棵树。我们的想法是识别包含多个分段主干的分段冠。其中一根树干链接到分段的树冠（最上层的树），其余的被视为地下树干。在树木体积的非线性混合效应模型中，冠状分割树的高度和林下树干的直径被用作预测变量。在30 m×30 m验证图级别上，ITC和HTD的体积预测的％RMSE和％MD值分别为22.2％，-13.3％和18.8％，-8.3％。我们得出的结论是，当使用UALS数据时，建议的HTD方法可提高管理的北方森林中ITC的准确性。