基于标记控制分水岭算法的树冠高程模型分割

Canopy Height Model Segmentation Based on Marker Controlled Watershed Algorithm

树冠高程模型是树冠表面模型和数字高层模型的差值,是森林资源调查的重要数据源。笔者提出利用标记控制分水岭算法进行树冠高程模型分割的方法,首先用高斯滤波平滑树冠高程模型,对树冠边缘做初步检测,然后利用形态学重构方法进行树冠形状确认,在此基础上进行二值图像的距离变换和h-minima变换,并标记树冠顶部,最后对实验数据进行分水岭分割,实现单株木树冠边缘勾勒。误差分析表明,该方法能有效提高分割正确率,当结构元素和h-minima变换中的参数分别等于3、1.5时,树冠分割正确率为78.5%。

Canopy height model （CHM） is the difference between canopy surface model with digital elevation model, which is forest resources inventory important data. The author proposed CHM segmentation based on marker controlled watershed algorithm. Firstly, the author used the Gaussian filter smooth CHM, obtained the crown edge; secondly, the author affirmed the crown shape by morphological reconstruction, and then exploited distance and h-minima transform to label treetop; lastly, used watershed algorithm segment CHM and outlined the canopy edge. Experimental results were provided to illustrate this method, which could effectively improve the recognition accuracy, the segmentation accuracy was 78.5% when parameter b and h was 3 and 1.5, respectively.