基于局域最大值法单木位置探测的适宜模型研究

以凉水自然保护区为研究区域,基于机载激光雷达数据,采用动态窗口局部最大值法对郁闭度较高的针叶林进行单木位置自动提取。采用树冠高度模型(CHM)和树冠最大模型(CMM)配合两种动态窗口,即树高--树冠大小回归方程和该方程的95%预测下限来探测树冠顶点,用探测百分比、1∶1对应关系的单木个数、生产者精度和用户精度进行了精度评价。结果表明:利用CMM能够抑制树冠内部枝杈产生的错判现象;利用树高--树冠大小回归方程95%的预测下限做动态窗口,能够有效防止在局部最大值方法中产生的小树漏测现象。因此,利用CMM和95%的预测下限做动态窗口的局域最大值法有利于提高单木位置探测的精度,为密林中自动地探测单木位置提供依据。 

Suitable model of detecting the position of individual treetop based on local maximum method

Based on airborne laser scanning data, we employed local maximum method with variable window size to detect the positions of individual trees in high-density forest of Liangshui Nature Reserve in Heilongjiang Province. Two models, canopy height model ( CHM) and canopy maximum model ( CMM) , were applied to detect the treetop; tree height-crown size regression and its 95% lower predicting limit were used to adjust variable window size, and precisions were assessed by the number of detection percentage, 1:1 hits, producer's accuracy and user's accuracy. The results showed that: CMM could restrain commission error caused by branches within canopies; meanwhile, variable window size defined by 95% lower predicting limit of the regression could avoid omission error caused by small trees with local maximum method. Thus, treetop detection using CMM and 95% lower predicting limit of regression as window size in local maximum method could improve accuracy of measuring the position of individual tree and provide theoretical basis for automatically detecting individual trees in high-density forests.