基于形态学阈值标记分水岭算法的高分辨率影像单木树冠提取  

Extraction of High-resolution Images of Single Tree Crown Based on Watershed Algorithm with Morphological Threshold Mark

作　　者：郑鑫 王瑞瑞[1] 靳茗茗 

机构地区：[1]北京林业大学林学院,北京100083 [2]中国地质大学(北京)地球科学与资源学院,北京100083 [3]中国科学院大学,北京100049 [4]中国科学院地理科学与资源研究所,北京100101

出　　处：《中南林业调查规划》2017年第4期30-35,57,共7页Central South Forest Inventory and Planning

基　　金：中央高校基本科研业务费专项资金资助(YX2014-09);国家自然科学基金资助项目"基于条件随机场模型和森林三维形态结构的树种分类算法研究"(41201446)

摘　　要：随着遥感技术的不断发展,利用高空间分辨率遥感影像提取单木树冠,成为获取树冠信息的一种重要手段。结合数学形态学和最大类间方差法提取自适应的分割阈值,分别对前景和背景进行标记,以此构建了改进的分水岭分割方法。选取林区高分辨率的无人机影像为数据源,分别根据传统的分水岭方法和改进的分水岭方法进行实验分析,并从冠层面积、单木树冠分割的数量和质量方面进行精度评价。结果证明:基于形态学阈值标记的分水岭算法具有显著改善过分割的作用,树冠分割结果精度可达64.62%,表明该方法切实可行,且效果较好。At present, with the continuous development of the remote sensing technology, using the high spatial resolution remote sensing image to extract the single tree crown is becoming an important means that can obtain the canopy information. Aiming at the phenomenon of traditional watershed segmentation algorithm, the adaptive segmentation threshold was extracted by combining mathematical morphology and the maximum interclass variance method, and the foreground and background were marked respectively to construct the improved watershed segmentation method. We selected the high -resolution UAV images as the data source, according to the traditional watershed method and the improved watershed method, the experimental analysis was carried out. And the accuracy of the quantity and quality of the canopy dividing and the areas of single tree crowns were evaluated. The result showed that the watershed algorithm based on the morphological threshold mark had the effect of improving the phenomenon of over - segmentation. The accuracy of the crown segmentation could reach to 64. 62%, which indicates that this method is feasible and effective.

关 键 词：树冠提取 自适应阈值 图像分割 分水岭算法 数学形态学 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]