在森林资源调查中冠幅和胸径是重要的测树因子,自动获取冠幅和胸径值可以提高森林资源调查效率。以云南松为研究对象,基于无人机影像自动提取单木冠幅参数,拟合不同密度等级样地的单木冠幅和树冠面积与胸径的关系以估测单株胸径。首先利用标记控制分水岭分割算法对样地冠层高度模型(CHM)中的单株树冠进行分割,获取最大、最小冠幅和树冠面积,并与实测数据进行精度评价,然后将提取冠幅与树冠面积与实测胸径进行拟合,建立不同密度等级样地的一元回归模型和二元回归模型。结果表明:单木树冠分割准确率为86.26%,冠幅相对误差平均值为6.04%,冠幅面积的相对误差平均值为11.23%;在拟合的模型中,冠幅&树冠面积-胸径模型的拟合效果最好,决定系数均在0.7以上,该模型验证数据相对误差均不超过5%,符合A类森林资源调查胸径误差值低于5%的要求。提出的基于无人机影像提取冠幅及预测树木胸径的方法较为准确,可推动森林资源调查自动化发展。

Crown width and DBH are important tree measurement factors in forest resources survey.Automatic acquisition of crown width and DBH data can improve the efficiency of forest resources investigation.In this study,Pinus yunnanensis was selected as the research object.Based on UAV images,the parameters of individual tree crown width were extracted automatically,and the relationship between crown width,crown area and DBH of different density grade plots was fitted to predict the DBH value of trees.Firstly,the individual tree crown in the canopy height model (CHM) was segmented by marker controlled watershed segmentation algorithm,and the maximum and minimum crown width and crown area were obtained,and the accuracy was evaluated with the mea-sured data.Then,the extracted crown width and crown area were fitted with the measured DBH,and the univariate and binary regression models of each density grade plot were established.The results showed that:the accuracy rate of individual tree crown segmentation was 86.26%,the average relative error of crown width was 6.04%,and the average relative error of crown area was 11.23%;in the fitting model,the fitting effect of crown width & crown area DBH model was the best,the determination coefficient was above 0.7,the relative error of validation data of the model was less than 5%,which met the requirements of class A forest resources survey.The method of extracting crown width and predicting DBH of trees based on UAV image is more accurate,which can provide reference for automatic development of forest resources investigation.