松材线虫病遥感监测研究进展及方法述评

Research progress on remote sensing monitoring of pine wilt disease

该研究旨在综合述评松材线虫病遥感监测的历史及近年来的研究进展,并就当前研究和工作中产生的问题给出建议与展望,为相关管理部门、科研院所以及从业者提供技术参考和辅助决策依据。该研究以科学引文数据库(WoS)和CNKI检索并筛选后得到的文献为基础,系统梳理松材线虫病遥感监测的提出及发展;根据遥感监测的对象层次分类梳理了相关研究中使用的方法,就当前研究中存在的问题进行归纳和总结,最后给出了未来可能的工作及研究方向。该研究发现:1)2017年来松材线虫病遥感监测一直处于研究热点。2)松材线虫病遥感监测研究使用数据的运载平台大多为以无人机为代表的机载平台,而光谱类型以RGB和多光谱为主。3)松材线虫病遥感监测的粒度以单株为主,监测使用的病害类别体系繁多且不同类别体系间的关系模糊。4)机器学习和深度学习两类方法在松材线虫病遥感监测研究中占据垄断地位,但两类方法各有优势、互不取代。该研究认为遥感调查极大提高了松材线虫病疫情增量控制及存量消减工作的效率,但存在单一数据源难以满足大范围细粒度的监测需求、病害类别体系杂乱、数据集不统一不标准、缺乏长时序监测成果等方面的问题。该研究提出未来可以在空天数据融合、病害类别体系及数据集标准化和短周期长时序监测等3个方面进一步开展工作及研究,将有助于松材线虫病遥感监测的进一步实时化和智能化。

Pine Wilt Disease(PWD),a devastating pine tree disease,has caused a serious impact on the national biosecurity,ecological security,and forestry economy.In this study,a systematic review of the research progress was made on the history of remote sensing monitoring of PWD in recent years under the object level classification of remote sensing monitoring using the literature retrieved and screened by the Web of Science(WoS)and China National Knowledge Infrastructure(CNKI).Some suggestions and outlooks were also proposed for the existing problems,which could provide reference for the technical reference and auxiliary decision-making on forestry.It was found that:1)About 70%of the literature was published in the research field after 2017.It infers that the remote sensing monitoring of pine wood nematode has been a research hotspot in the past five years.2)From the viewpoint of the carrier platform,the satellite,airborne,and ground datasets accounted for 17.1%,75.6%,and 7.3%of the research data on the remote sensing monitoring of PWD,respectively.Particularly,there was the vast majority of airborne data represented by Unmanned Aerial Vehicles(UAV).From the viewpoint of data spectral type,44.0%,34.1%,17.1%,and 4.9%of the studies used RGB,multispectral,hyperspectral,and LIDAR data,respectively.Therefore,the RGB and multispectral datasets were dominated in the remote sensing monitoring of PWD.3)Single plants were mainly used as the granularity of remote sensing monitoring of PWD.The diseased trees were classified into the two,three,four,five,and six categories,accounting for 53%,23%,15%,6%,and 3%,respectively.There were diverse category systems with vague relationships between them.4)Machine learning and deep learning dominated the remote sensing monitoring of PWD.Furthermore,machine learning and deep learning shared their own advantages and fail to replace each other.Furthermore,the aerospace remote sensing survey with the UAV and satellite sensors as the data sources greatly improved the efficiency of PWD epidemic increment control and stock abatement work.However,the following challenges remained:1)A single data source cannot fully meet the harsh requirement of large-scale and fine-grained monitoring in recent years.2)Disorganized disease classification systems led to the irregularity and specification of data for machine learning and deep learning.3)It is still lacking in long-term series monitoring with the high-time resolution.Finally,three recommendations were proposed for the future real-time and intelligent remote sensing monitoring of PWD:①To explore the satellite and aerial data fusion for the large-scale and fine-grained disease monitoring;②To clarify the disease monitoring category system,and then to construct the relevant spectral library and sample library datasets;③To develop the high-frequency and long-time series remote sensing monitoring products for a general release mechanism for the PWD.