基于多尺度标记优化分水岭方法的油茶冠幅提取

针对尺度对地物空间结构的限制以及传统分水岭分割易产生树冠过分割等问题,选择长沙县明月村油茶基地为研究区,提出一种基于多尺度标记优化分水岭分割油茶树冠的方法。首先使用高分辨率无人机影像采集图像,分析影像特征,构建油茶分类体系,提取油茶林分布区域。其次,运用多尺度区域迭代增长方法提取树冠标记,将标记应用于多阈值尺度的分水岭变换,并结合Johnson指数选取树冠标记增长和分水岭阈值的最优尺度,实现油茶单木的准确识别。结果表明:多尺度标记优化分水岭方法在分离油茶单木时,树冠面积提取值与目视解译参考值的相对误差为9.4%;单木总体识别精度为89.4%,相对于传统的分水岭分割方法精度提高了34.8%;通过Johnson指数确定的最优迭代增长尺度为20,分水岭分割阈值尺度为85,对比不同尺度组合下的油茶冠幅提取结果,最优尺度下的油茶冠幅提取精度最高(R²=0.75)。多尺度标记优化分水岭方法能较准确地分离油茶树冠,将该方法应用于无人机影像树冠分割,可有效提高经济林调查的效率。

Extraction of Camellia oleifera crown width based on the method of optimized watershed with multi-scale markers

In view of the limitation of scale on the spatial structure of ground objects and the problem that traditional watershed segmentation tends to produce crown over-segmentation, we proposed a segmentation method of Camellia oleifera crown based on the optimized watershed with multi-scale markers, with the C. oleifera base in Mingyue Village of Changsha County as the research object. Firstly, the high-resolution unmanned aerial vehicle (UAV) was used to collect images. The image features were analyzed to construct the classification system of C. oleifera, and the distribution area of C. oleifera was extracted. After being extracted by multi-scale region iterative growth, the crown markers were applied to the multi-threshold scale watershed transformation. Combined with Johnson index, the optimal scale of crown marker growth and watershed threshold was used to realize the accurate identification of individual trees. The results showed that the relative error between the method of optimized watershed with multi-scale markers and the visual interpretation of the reference value of tree-crown was 9.4% for the separation of individual trees. The overall identification accuracy of each tree was 89.4%, which was 34.8% higher than that of the traditional watershed segmentation method. The optimal iterative growth scale obtained by Johnson index was 20, while the thre-shold scale of watershed segmentation was 85. Compared with the results of different scale combinations, the crown extraction accuracy under the optimal scale was the highest (R²=0.75). The method of optimized watershed with multi-scale markers could accurately separate C. oleifera crown. Applying this method to UAV image crown segmentation could effectively improve the efficiency of economic forest investigation.