热带地理 ›› 2019, Vol. 39 ›› Issue (4): 553-561.doi: 10.13284/j.cnki.rddl.003148

• 专刊:无人机在生态学和地理学中的应用 • 上一篇    下一篇

基于两种无人机航拍影像的林窗和林冠提取分析

王 悦1,2,练琚愉2,张昭臣3,胡健波4,杨 骥5,李 勇5,叶万辉2   

  1. (1. 中国科学院大学,北京 100049;2. 中国科学院华南植物园,广州 510650;3. 华东师范大学 生态与环境科学学院,上海 200241; 4. 交通运输部天津水运工程科学研究所,天津 300456;5. 广州地理研究所,广州 510070)
  • 出版日期:2019-07-10 发布日期:2019-07-10
  • 通讯作者: 练琚愉(1976—)女,广东人,副研究员,博士,研究方向为保育生态学,(E-mail)lianjy@scbg.ac.cn。
  • 作者简介:王悦(1991—),女,新疆人,硕士,研究方向为保育生态学,(E-mail)wangyue@scbg.ac.cn;
  • 基金资助:

    国家重点研发计划项目(2017YFC0505802);中国科学院B类先导科技专项项目(XDB3103);广东省科技计划项目(2017B010117008)

Forest Plots Gap and Canopy Structure Analysis Based on Two UAV Images

Wang Yue1,2, Lian Juyu2, Zhang Zhaochen3, Hu Jianbo4, Yang Ji5, Li Yong5 and Ye Wanhui2   

  1. (1. University of Chinese Academy of Sciences, Beijing 100049, China; 2. South China Botanical Garden, Chinese Academy of Sciences, Guangzhou 510650, China; 3. School of Ecological and Environmental Sciences, East China Normal University, Shanghai 200241, China; 4. Tianjin Research Institute of Water Transport Engineering, Ministry of Transport, Tianjin 300456, China; 5. Guangzhou Institute of Geography, Guangzhou 510070, China)
  • Online:2019-07-10 Published:2019-07-10

摘要:

森林冠层和林窗的结构及其时空变化是理解森林生态系统格局、动态变化过程的重要基础。在当前生物多样性监测倍受关注的契机下,如何以合适的手段准确描述林窗面积、分布等特征,并与森林固定样地监测数据有效地结合,更好地回答群落构建的理论问题,使森林群落物种多样性维持机制得到更全面的认识,是目前亟待解决的问题。以鼎湖山南亚热带常绿阔叶林20 hm2固定监测样地为研究对象,基于不同遥感影像提取方法对其林窗和林冠表层数据进行提取分析。结果表明:基于监督分类的提取方法适合RGB波段航片林窗的提取,在林窗分类中,应首先确定林窗高度、边界木与最小面积,不同分类方法差异主要表现在林冠分类中,林窗分类生产者精度和用户精度表现都较为一致。无人机航拍识别率受地形因素影响较大,在地形复杂林地应按坡度分区域进行飞行以降低误差。相对于地面调查,MD4-1000无人机航片的林窗识别率为98.7%;大疆Phantom4无人机航片的林窗识别率为72.3%,影像后期处理数据量小,同样适用于森林林窗定量研究,符合生态学、林业等从业人员对大型样地林窗长期监测的要求。无人机航拍南亚热带森林物种识别难度较大,基于MD4-1000无人机搭载的高分辨率相机,在地势平缓区域优选的4 hm2样地中可识别林冠表层物种数17种,共2 706个个体。搭载高分辨率无人机在降低飞行高度的基础上可进行部分物种识别。应用无人机近地面遥感对森林固定样地进行林冠监测,可为后期群落构建研究提供数据基础,有望从新的研究角度探讨森林群落物种多样性维持机制。

关键词: 无人机, 南亚热带常绿阔叶林, 鼎湖山, 林窗, 物种识别

Abstract:

Forest canopy plays a significant role on community biodiversity maintenance, which is the key eco-boundary of forest atmosphere interactions. As an indicator of canopy dynamic, gap is one of the most important factors in maintaining the long-term transformation of ecosystems. In recent times, biodiversity monitoring has been the focus of much attention and study. As such, the problem of how to accurately describe gap features such as size and distribution requires urgent resolution. Furthermore, it is necessary to combine these features with monitoring data from permanent plots in forests in order to bring about solutions for issues relevant to community construction. These solutions will also help researchers achieve a better understanding of the maintenance system of community species diversity in forests. In this study, a practical canopy gap monitoring system was formed on the basis of different image extraction methods using lightweight Unmanned Aerial Vehicles (UAV) and a Geographic Information System (GIS). A 20 ha permanent monitoring plot in Dinghushan, classified as subtropical forest stand in South China, was selected as the data source. The results obtained from different gap extraction methods were scientifically analyzed after a strict classification. Results indicated that the red, green, and blue (RGB) band image classification was applied as the method of extracting remotely sensed images in the monitoring system built in the study. The extracted results were significantly similar to precise field measures. We emphasized penetrability of canopy as a whole in order to quantify the concept depths of the forest gap in subtropical forest stands; we confirmed that the gap could be established in only the lowest height of the canopy. The gap accuracy of supervised classification based on the DJI Phantom UAV was 72.3%, a value lower than the 98.7% attributed to the MD4-1000 UAV. The MD4-1000 is able to fly at specific heights according to geographic states and retrieve relevant tree height data; however, the cost of its use need to be considered. The differences in data obtainment between extraction and actual measurement will also increase while the gradient of plot rises. Therefore, the DJI Phantom UAV is considered suitable for large plot gap extraction due to its accuracy and high efficiency and despite its low-resolution ratio and decreased mission function. Nonetheless, a total of 2 706 individuals and 17 species were identified by MD4-1000 UAV. A high-resolution drone is available for partial species identification based on reduced flight altitude. This paper therefore indicates that monitoring forest canopies of permanent plots by using UAV on the basis of near-ground remote sensing is able to provide a database for study in community assembly. It is expected that species diversity maintenance can also be investigated by the inclusion of variants in our study perspective. t is expected that the inclusion of variants in our study perspective could provide further database for study in forest community assembly. Hopefully species diversity maintenance could be investigated by new perspective of near-ground remote sensing.

Key words: Unmanned Aerial Vehicles, subtropical evergreen broad-leaved forest, Dinghushan, forest gap, species identification