Segmentation of individual tree crowns in colour aerial photographs using region growing supported by fuzzy rules
Segmentation of individual tree crowns in colour aerial photographs using region growing supported by fuzzy rules
复制标题
DOI：
10.1139/x03-062
复制
发表时间：
2003-08-01
期刊：
影响因子：
--
通讯作者：
Erikson, M
中科院分区：
文献类型：
--
作者：
Erikson, M
提出了一种基于区域生长的航空照片树冠分割方法。通过在空间域和颜色域中同时使用是否包含像素的决策函数，在分割结果中保留了树冠的不规则轮廓。因此，轮廓信息随后可用于树种分类。评估每棵树的一组可能的候选区域，并根据度量选择最佳区域。该方法在瑞典中部的大量高空间分辨率航空图像样本上进行了评估。图像材料中包含近 15 公顷的天然再生北方森林。平均而言，73%的树冠识别结果与人工勾画的相应结果一致。 30张测试图像的茎总数估计为手动计数值的93%。
A method based on region growing for segmentation of tree crowns in aerial photographs is presented. By using a decision function, for including a pixel or not, in the spatial domain and in the colour domain simultaneously, the irregular contour of the tree crowns is kept in the segmentation result. Thus, contour information may be subsequently used for tree species classification. A set of possible candidate regions for each tree is evaluated, and the best region, according to a measure, is selected. The method is evaluated on a large sample of high spatial resolution aerial images in central Sweden. Almost 15 ha of natural regenerated boreal forests are included in the image material. On average, the identification of 73% of the tree crowns is in agreement with the corresponding results from the manual delineations. The estimation of the total number of stems from the 30 test images is 93% of the value obtained from the manual count.