From Forests to Cities: How Does Hyperspectral Imaging Enable Tree Species Identification?

高光譜成像技術(shù)在樹(shù)種識(shí)別領(lǐng)域的應(yīng)用日益廣泛，它通過(guò)捕捉樹(shù)木在多個(gè)窄波段上的光譜信息，實(shí)現(xiàn)對(duì)樹(shù)種的精確分類和識(shí)別，在森林資源管理、城市綠化規(guī)劃和生態(tài)環(huán)境保護(hù)等方面具有重要意義，為相關(guān)工作提供了關(guān)鍵技術(shù)支撐。

下面是高光譜成像技術(shù)在樹(shù)種識(shí)別的應(yīng)用場(chǎng)景。

Hyperspectral imaging technology is increasingly being applied in the field of tree species identification. By capturing spectral information from trees across multiple narrow bands, it enables accurate classification and identification of species. This technology plays a significant role in forest resource management, urban greening planning, and ecological environment protection, providing critical technical support for related tasks.

Below are the application scenarios of hyperspectral imaging technology in tree species identification.

1. 森林資源調(diào)查與監(jiān)測(cè) / Forest Resource Inventory and Monitoring

·樹(shù)種分類與分布：高光譜數(shù)據(jù)可以用于識(shí)別和分類森林中的不同樹(shù)種，生成樹(shù)種分布圖，為森林資源管理提供基礎(chǔ)數(shù)據(jù)。

在巴西大西洋森林的研究中，研究者結(jié)合無(wú)人機(jī)高光譜數(shù)據(jù)與激光雷達(dá)（LiDAR）數(shù)據(jù)，對(duì)8種上層樹(shù)冠樹(shù)種進(jìn)行分類，通過(guò)主成分分析（PCA）處理所有特征后，分類總體精度達(dá)到76%，為該退化森林的物種分布監(jiān)測(cè)提供了有效數(shù)據(jù)支撐。

·Tree Species Classification and Distribution: Hyperspectral data can be used to identify and classify different tree species in forests, generating species distribution maps that serve as foundational data for forest resource management.

In a study of the Atlantic Forest in Brazil, researchers combined UAV-based hyperspectral data with LiDAR data to classify eight canopy tree species. After processing all features using Principal Component Analysis (PCA), an overall classification accuracy of 76% was achieved, providing effective data support for monitoring species distribution in this degraded forest.

各種樹(shù)的平均光譜 / Mean spectra for each tree species

·森林健康評(píng)估：通過(guò)分析樹(shù)木的光譜特征，可以評(píng)估樹(shù)木的生長(zhǎng)狀況和健康程度，及時(shí)發(fā)現(xiàn)病蟲(chóng)害和環(huán)境脅迫，為森林保護(hù)提供預(yù)警信息。

中國(guó)地質(zhì)調(diào)查局在湖北宜城的研究中，采用無(wú)人機(jī)高光譜數(shù)據(jù)（400~1000nm，270個(gè)光譜波段），結(jié)合歸一化植被指數(shù)（NDVI）、類胡蘿卜素反射指數(shù)（CRI）和水波段指數(shù)（WBI），構(gòu)建“寬帶綠度指數(shù)-葉綠素指數(shù)-冠層含水量/光合能力指數(shù)"的綜合評(píng)估體系，實(shí)現(xiàn)了森林樹(shù)木健康狀況的定性與定量評(píng)估，結(jié)果與實(shí)地觀測(cè)及假彩色合成圖像特征高度一致。

·Forest Health Assessment: By analyzing the spectral characteristics of trees, their growth conditions and health status can be evaluated, enabling timely detection of pests, diseases, and environmental stressors, thereby offering early warning information for forest protection.

In a study conducted by the China Geological Survey in Yicheng, Hubei, UAV-based hyperspectral data (400-1000nm, 270 spectral bands) was used in combination with vegetation indices such as NDVI, CRI, and WBI to construct a comprehensive evaluation system based on "broadband greenness index–chlorophyll index–canopy water content/photosynthetic capacity index." This system achieved both qualitative and quantitative assessments of forest tree health, with results highly consistent with field observations and false-color composite imagery.

(a) 真彩色影像與樹(shù)種識(shí)別分類；(b) 假彩色影像與健康評(píng)估

(a) True color image and tree species recognition class; (b) False color image and health assessment

·生物多樣性研究：高光譜數(shù)據(jù)可以用于研究森林生態(tài)系統(tǒng)的生物多樣性，了解不同樹(shù)種的生態(tài)功能和相互關(guān)系，為生態(tài)保護(hù)提供科學(xué)依據(jù)。

·Biodiversity Research: Hyperspectral data can be applied to study biodiversity in forest ecosystems, helping to understand the ecological functions and interrelationships of different tree species, thereby providing a scientific basis for ecological conservation.

·林木生長(zhǎng)參數(shù)反演：利用高光譜數(shù)據(jù)可以反演林木的葉面積指數(shù)、生物量等生長(zhǎng)參數(shù)，為林木生長(zhǎng)模型的建立和優(yōu)化提供數(shù)據(jù)支持。

在東北針闊混交林研究中，研究者通過(guò)高光譜數(shù)據(jù)提取植被指數(shù)，結(jié)合LiDAR獲取的樹(shù)高、冠幅等結(jié)構(gòu)參數(shù)，實(shí)現(xiàn)了林木葉面積指數(shù)和生物量的精準(zhǔn)反演，為該區(qū)域精準(zhǔn)林業(yè)中林木生長(zhǎng)模型優(yōu)化提供了關(guān)鍵數(shù)據(jù)。值得注意的是，在這項(xiàng)研究中，高光譜成像儀和LiDAR是分別掛載在不同的無(wú)人機(jī)上的。

·Inversion of Tree Growth Parameters: Hyperspectral data can be used to invert growth parameters such as leaf area index and biomass, supporting the establishment and optimization of tree growth models.

In a study on mixed coniferous-broadleaf forests in Northeast China, researchers extracted vegetation indices from hyperspectral data and combined them with structural parameters (e.g., tree height and crown width) obtained from LiDAR to achieve accurate inversion of leaf area index and biomass. This provided key data for optimizing tree growth models in precision forestry in the region. It is worth noting that in this study, the hyperspectral imager and LiDAR were mounted on different UAVs.

樹(shù)種專題圖 / Thematic map of tree species

2. 城市綠化規(guī)劃與管理 / Urban Greening Planning and Management

·城市樹(shù)種識(shí)別與分布：高光譜圖像可以用于識(shí)別城市中的樹(shù)種，了解城市綠化的樹(shù)種構(gòu)成和分布情況，為城市綠化規(guī)劃提供參考。

·城市樹(shù)木健康監(jiān)測(cè)：通過(guò)分析城市樹(shù)木的光譜特征，可以評(píng)估城市樹(shù)木的生長(zhǎng)狀況和健康程度，及時(shí)發(fā)現(xiàn)病蟲(chóng)害和環(huán)境脅迫，為城市樹(shù)木的養(yǎng)護(hù)管理提供指導(dǎo)。

香港理工大學(xué)一團(tuán)隊(duì)利用高光譜圖像對(duì)城市樹(shù)種進(jìn)行了識(shí)別分類，2018年11月至2019年10月期間，在不同季節(jié)對(duì)19個(gè)樹(shù)種的75棵城市樹(shù)木進(jìn)行了圖像采集，深度神經(jīng)網(wǎng)絡(luò)方法在物種識(shí)別中達(dá)到了85%~96%的準(zhǔn)確率。不同物種對(duì)健康狀況表現(xiàn)出不同的光譜響應(yīng)。

·Urban Tree Species Identification and Distribution: Hyperspectral imagery can be used to identify tree species in urban areas, helping to understand the composition and distribution of species in urban greening, thus providing references for urban greening planning.

·Urban Tree Health Monitoring: By analyzing the spectral characteristics of urban trees, their growth conditions and health status can be assessed, enabling timely detection of pests, diseases, and environmental stressors, thereby guiding maintenance and management efforts.

A team at The Hong Kong Polytechnic University used hyperspectral imagery to identify and classify urban tree species. From November 2018 to October 2019, images of 75 urban trees from 19 species were collected across different seasons. Deep neural network methods achieved an accuracy of 85%–96% in species identification. Different species exhibited distinct spectral responses to health conditions.

(a-d)為原始圖像；(e-h)為對(duì)應(yīng)的掩蔽后圖像

Typified examples of masking canopies and homogenous regions: (a-d) are original images; (e-h) are corresponding masked images.

各輪實(shí)地?cái)?shù)據(jù)采集中的不同樹(shù)種平均冠層光譜特征，樹(shù)種分別為：(a) 相思樹(shù)（樣本量N=6）；(b) 大葉合歡（N=3）；(c) 白楸（N=5）；(d) 榕樹(shù)（N=3）

Mean canopy spectral signature of different species in each round of in-situ data acquisition, the species are: (a) Acacia confuse (N = 6); (b) Albizia lebbeck (N = 3); (c) Mallotus paniculatus; (N = 5); (d) Ficus macrocarpa (N = 3). N indicates the number of tree samples for the corresponding species;

高光譜成像技術(shù)為樹(shù)種識(shí)別提供了高效、精確的技術(shù)方案，它可以減少人工調(diào)查的工作量、獲取精細(xì)信息、為森林資源管理、城市綠化規(guī)劃和生態(tài)環(huán)境保護(hù)提供科學(xué)依據(jù)和決策支持，應(yīng)用前景廣闊。

隨著技術(shù)發(fā)展，它將進(jìn)一步助力林業(yè)與生態(tài)領(lǐng)域的可持續(xù)發(fā)展，持續(xù)發(fā)揮核心支撐作用。

作為高光譜的供應(yīng)商，愛(ài)博能提供全面的產(chǎn)品線，包括全波段的高光譜相機(jī)、無(wú)人機(jī)載高光譜成像系統(tǒng)、便攜式、高光譜實(shí)驗(yàn)室和顯微高光譜。歡迎垂詢！

Hyperspectral imaging technology provides an efficient and accurate technical solution for tree species identification. It reduces the workload of manual surveys, captures detailed information, and offers scientific basis and decision-making support for forest resource management, urban greening planning, and ecological environment protection. Its application prospects are broad.

With technological advancements, it will further contribute to the sustainable development of forestry and ecology, continuing to play a core supporting role.

As a supplier of hyperspectral solutions, ExponentSci provides a comprehensive product line, including full-band hyperspectral cameras, UAV-mounted hyperspectral imaging systems, portable systems, hyperspectral laboratories, and micro-hyperspectral imagers. Welcome to inquire!

案例來(lái)源 / Sources：

1. Zhong, H., Lin, W., Liu, H., Ma, N., Liu, K., Cao, R., Wang, T., & Ren, Z. (2022). Identification of tree species based on the fusion of UAV hyperspectral image and LiDAR data in a coniferous and broad-leaved mixed forest in Northeast China. Frontiers in Plant Science, 13, 964769.

2. Martins-Neto, R. P., Tommaselli, A., Imai, N., Honkavaara, E., Miltiadou, M., Moriya, E., & David, H. (2023). Tree species classification in a complex Brazilian tropical forest using hyperspectral and LiDAR data. Forests, 14(5), 945.

3. Zeng, G., Xu, J., Zhang, W., & Wang, B. (2023). Tree species identification and health assessment of forest sample plots based on UAV hyperspectral remote sensing technology. Journal of Physics: Conference Series, 2621(1), 012001.

4. Abbas, S., Peng, Q., Wong, M. S., Li, Z., Wang, J., Ng, K. T. K., Kwok, C. Y. T., & Hui, K. K. W. (2021). Characterizing and classifying urban tree species using bi-monthly terrestrial hyperspectral images in Hong Kong. ISPRS Journal of Photogrammetry and Remote Sensing, 177, 204–216.