ispltaИзвестия Санкт-Петербургской лесотехнической академииIzvestia Sankt-Peterburgskoj lesotehniceskoj akademii2079-43042658-5871СПбГЛТУ10.21266/2079-4304.2025.255.89-102isplta-594Research ArticleЛЕСНОЕ ХОЗЯЙСТВОFORESTRYМашинное обучение без учителя на основе вегетационных индексов и спектров отражения в задаче классификации деревьевUnsupervised machine learning based on vegetation indices and reflectance spectra for tree classificationhttps://orcid.org/0009-0009-4241-3798СтешинИ. С.SteshinI. S.

Стешин Илья Сергеевич – младший научный сотрудник

424000, пл. Ленина, д. 3, г. Йошкар-Ола, Республика Марий Эл

Steshin Ilya S. – junior researcher

124000, Lenin Sq. 3. Yoshkar-Ola. Mari El

ilyasteshinm@yandex.ruhttps://orcid.org/0009-0000-2365-4857ПетуховИ. В.PetukhovI. V.

Петухов Игорь Валерьевич – профессор кафедры проектирования и производства электронно-вычислительных средств

424000, пл. Ленина, д. 3, г. Йошкар-Ола, Республика Марий Эл

Petukhov Igor V. – DSc (Technical), Professor of the Department of Design and Production of Electronic Computing Tools

124000, Lenin Sq. 3. Yoshkar-Ola. Mari El

PetuhovIV@volgatech.netПоволжский государственный технологический университетРоссияVolga State Technological UniversityRussian Federation202513122025025589102Copyright © Стешин И.С., Петухов И.В., 20252025Стешин И.С., Петухов И.В.Steshin I.S., Petukhov I.V.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://izvestiya-lta.spbftu.ru/jour/article/view/594

Статья посвящена применению метода машинного обучения k-means и вегетационных индексов для классификации древесных пород на основе материалов съемки с БПЛА. Вегетационные индексы анализируют спектральные характеристики растений, что особенно полезно для классификации схожих видов деревьев, поскольку они имеют разные спектральные сигнатуры. Исследование проведено на двух лесных участках с распределением деревьев «сосна – береза» и «сосна – ель». Данные были собраны с 5 спектральных датчиков – cинего (470 нм), зеленого (560 нм), красного (665 нм), дальнего красного (720 нм) и инфракрасного (840 нм). Детекция крон деревьев была произведена по RGB изображениям, полученным с помощью спектральных датчиков и модели DeepForest. Далее внутри границы детектированных крон деревьев данные со спектральных датчиков обрабатывались для расчета вегетационных индексов каждого дерева. Пиксели по каждому дереву из матричного вида были разложены в вектор, отсортированный по возрастанию. Из каждого вектора были рассчитаны следующие параметры: минимальное, максимальное, среднее, медианное значения, а также 25-й и 75-й квантили и энтропия Шеннона. После чего применялся метод понижения размерности PCA и кластеризации k-means для разделения типов деревьев. Результаты показали, что в случае с сосной и елью точность классификации превышала 80%, тогда как для сосны и березы метрики были менее успешными, что может быть связано с сезонными изменениями индексов. В обоих случаях лучшими индикаторами оказались DVI, демонстрируя высокую эффективность для определенных лесных покрытий.

The article is devoted to the use of the k-means machine learning method and vegetation indices for the classification of tree species based on UAV survey materials. Vegetation indices analyze the spectral characteristics of plants, which is especially useful for distinguishing similar tree species, as they exhibit different spectral signatures. The study was conducted on two forest plots with tree distributions of pine-birch and pine-spruce. Data were collected from five spectral sensors: Blue (470 nm), Green (560 nm), Red (665 nm), Far Red (720 nm), and Infrared (840 nm). Tree crown detection was performed using RGB images captured by spectral sensors and the DeepForest model. Within the boundaries of the detected crowns, spectral sensor data were processed to calculate the vegetation indices for each tree. Pixels for each tree from the matrix view were decomposed into a vector, sorted in ascending order. The following parameters were calculated from each vector: minimum, maximum, mean, median values, as well as the 25th and 75th percentiles and Shannon entropy. Dimensionality reduction via PCA and k-means clustering were then applied to differentiate tree types. The results showed that for pine and spruce, classification accuracy exceeded 80%, whereas for pine and birch, the metrics were less successful, possibly due to seasonal changes in the indices. In both cases, DVI proved to be the most effective indicator, demonstrating high efficiency for certain forest covers.

машинное обучениетаксация деревьеввегетационные индексыБПЛАmachine learningtree identificationvegetation indicesUAVИсследование выполнено при поддержке Фонда содействия развитию малых форм предприятий в научно-технической сфереReferencesГеоскан Gemini Мультиспектр. URL: https://www.geoscan.ru/ru/products/gemini-m (дата обращения: 05.05.2024).Barandas M., Folgado D., Fernandes L., Santos S., Abreu M., Bota P., Liu Hui, Schultz T., Gamboa H. TSFEL: Time Series Feature Extraction Library. SoftwareX, 2020, vol. 11, art. no. 100456.Михайличенко А.А. Аналитический обзор методов оценки качества алгоритмов классификации в задачах машинного обучения // Вестник АГУ. 2022. Вып. 4(311). С. 52–59.Chehreh B., Moutinho A., Viegas C. Latest Trends on Tree Classification and Segmentation Using UAV Data–A Review of Agroforestry Applications. Remote Sens (Basel), 2023, vol. 15, iss. 9, art. no. 2263.Barandas M., Folgado D., Fernandes L., Santos S., Abreu M., Bota P., Liu Hui, Schultz T., Gamboa H. TSFEL: Time Series Feature Extraction Library // SoftwareX. 2020. Vol. 11. Art. no. 100456.Crutchfield J.P., Feldman D.P. Regularities Unseen, Randomness Observed: Levels of Entropy Convergence. 2001. URL: https:.arxiv.org/abs/cond-mat/0102181 (accessed: 14.09.2024)Chehreh B., Moutinho A., Viegas C. Latest Trends on Tree Classification and Segmentation Using UAV Data–A Review of Agroforestry Applications // Remote Sens (Basel). 2023. Vol. 15, iss. 9. Art. no. 2263.Fassnacht F.E., Latifi H., Stereńczak K., Modzelevska A., Lefsky M., Waser L.T., Straub C., Ghosh A. Review of studies on tree species classification from remotely sensed data. Remote Sens Environ., 2016, vol. 186, pp. 64–87.Crutchfield J.P., Feldman D.P. Regularities Unseen, Randomness Observed: Levels of Entropy Convergence. 2001. URL: https://arxiv.org/abs/cond-mat/0102181 (дата обращения: 14.09.2024).Geoscan Gemini Multispectral. URL: https://www.geoscan.ru/ru/products/gemini-m (accessed May 05, 2024).Fassnacht F.E., Latifi H., Stereńczak K., Modzelevska A., Lefsky M., Waser L.T., Straub C., Ghosh A. Review of studies on tree species classification from remotely sensed data // Remote Sens Environ. 2016. Vol. 186. P. 64–87.Hamunyela E., Verbesselt J., Herold M. Using spatial context to improve early detection of deforestation from Landsat time series. Remote Sens Environ., 2016, vol. 172, pp. 126–138.Hamunyela E., Verbesselt J., Herold M. Using spatial context to improve early detection of deforestation from Landsat time series // Remote Sens Environ. 2016. Vol. 172. P. 126–138.Harris C.R., Millman J., van der Walt S.J., Gommers R., Virtanen P., Cournapeau D., Wieser E., Taylor J., Berg S., Smith N.J., Kern R., Picus M., Hoyer S., van Kerkwijk M.H., Brett M., Haldane A., del Río J.F., Wiebe M., Peterson P., Gérard-Marchant P., Sheppard K., Reddy T., Weckesser W., Abbasi H., Gohlke H., Oliphant T.E. Array programming with NumPy. Nature, 2020, vol. 585, iss. 7825, pp. 357–362.Harris C.R., Millman J., van der Walt S.J., Gommers R., Virtanen P., Cournapeau D., Wieser E., Taylor J., Berg S., Smith N.J., Kern R., Picus M., Hoyer S., van Kerkwijk M.H., Brett M., Haldane A., del Río J.F., Wiebe M., Peterson P., Gérard-Marchant P., Sheppard K., Reddy T., Weckesser W., Abbasi H., Gohlke H., Oliphant T.E. Array programming with NumPy // Nature. 2020. Vol. 585, iss. 7825. P. 357–362.Huete A., Didan K., Miura T., Rodriguez E.P., Gao X., Ferreira L.G. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens Environ., 2002, vol. 83, iss. 1–2, pp. 195–213.Huete A., Didan K., Miura T., Rodriguez E.P., Gao X., Ferreira L.G. Overview of the radiometric and biophysical performance of the MODIS vegetation indices // Remote Sens Environ. 2002. Vol. 83, iss. 1–2. P. 195–213.Jollife I. T., Cadima J. Principal component analysis: A review and recent developments. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 2016, vol. 374, art. no. 2065.Jollife I.T., Cadima J. Principal component analysis: A review and recent developments // Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences. 2016. Vol. 374. Art. no. 2065.Macqueen J. Some methods for classification and analysis of multivariate observations. Berkeley Symp. on Math. Statist. and Prob., 1967, pp. 281–297.Macqueen J. Some methods for classification and analysis of multivariate observations // Berkeley Symp. on Math. Statist. and Prob. 1967. P. 281–297.Mikhaylichenko A.A. Analytical review of methods for assessing the quality of classification algorithms. ASU Bulletin, 2022, iss. 4(311), pp. 52–59. (In Russ.)Nevalainen O., Honkavaara E., Tuominen S., Viljanen N., Hakala T., Yu Xiaowei, Hyyppä J., Saari H., Pölönen I., Imai N.N., Tommaselli A.M.G. Individual tree detection and classification with UAV-Based photogrammetric point clouds and hyperspectral imaging // Remote Sens (Basel). 2017. Vol. 9, iss. 3. Art. no. 185.Nevalainen O., Honkavaara E., Tuominen S., Viljanen N., Hakala T., Yu Xiaowei, Hyyppä J., Saari H., Pölönen I., Imai N.N., Tommaselli A.M.G. Individual tree detection and classification with UAV-Based photogrammetric point clouds and hyperspectral imaging. Remote Sens (Basel), 2017, vol. 9, iss. 3, art. no. 185.Onishi M., Ise T. Automatic classification of trees using a UAV onboard camera and deep learning. 2018. URL: https://arxiv.org/abs/1804.10390 (дата обращения: 14.09.2024)Onishi M., Ise T. Automatic classification of trees using a UAV onboard camera and deep learning. 2018. URL: https:.arxiv.org/abs/1804.10390 (accessed: 14.09.2024)Pedregosa F., Varoquaux G., Gramfort A., Michel V., Thirion B., Grisel O., Blondel M., Prettenhofer P., Weiss R., Dubourg V., Vanderplas J., Passos A., Cournapeau D., Brucher M., Perrot M., Duchesnay É. Scikit-learn: Machine Learning in Python // Journal of Machine Learning Research. 2011. Vol. 12. P. 2825–2830.Pedregosa F., Varoquaux G., Gramfort A., Michel V., Thirion B., Grisel O., Blondel M., Prettenhofer P., Weiss R., Dubourg V., Vanderplas J., Passos A., Cournapeau D., Brucher M., Perrot M., Duchesnay É. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research, 2011, vol. 12, pp. 2825–2830.Pettorelli N., Vik J.O., Mysterud A., Gaillard J.-M., Tucker S.J., Stenseth N.C. Using the satellite-derived NDVI to assess ecological responses to environmental change // Trends Ecol Evol. 2005. Vol. 20, iss. 9. P. 503–510.Pettorelli N., Vik J.O., Mysterud A., Gaillard J.-M., Tucker S.J., Stenseth N.C. Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends Ecol Evol., 2005, vol. 20, iss. 9, pp. 503–510.Piekarski P., Zwoliński Z. Temporal variation in vegetation indexes for pine and beech stands during the vegetation season, Szczecin Lowland, Poland // Quaestiones Geographicae. 2014. Vol. 33, iss. 3. P. 131–143.Piekarski P., Zwoliński Z. Temporal variation in vegetation indexes for pine and beech stands during the vegetation season, Szczecin Lowland, Poland. Quaestiones Geographicae, 2014, vol. 33, iss. 3, pp. 131–143.Rautiainen M., Nilson T., Lükk T. Seasonal reflectance trends of hemiboreal birch forests // Remote Sens Environ. 2009. Vol. 113, iss. 4. P. 805–815.Rautiainen M., Nilson T., Lükk T. Seasonal reflectance trends of hemiboreal birch forests. Remote Sens Environ., 2009, vol. 113, iss. 4, pp. 805–815.Virtanen P., Gommers R., Oliphant T.E., Haberland M., Reddy T., Cournapeau D., Burovski E., Peterson P., Weckesser W., Bright J., van der Walt S.J., Brett M., Wilson J., Millman K.J., Mayorov N., Nelson A.R.J., Jones E., Kern R., Larson E., Carey C.J., Polat I., Feng Yu, Moore E.W., VanderPlas J., Laxalde D., Perktold J., Cimrman R., Henriksen I., Quintero E.A., Harris C.R., Archibald A.M., Ribeiro A.H., Pedregosa F., van Mulbregt P., SciPy 1.0 Contributors. SciPy 1.0: fundamental algorithms for scientific computing in Python // Nat. Methods. 2020. Vol. 17, iss. 3. P. 261–272.Virtanen P., Gommers R., Oliphant T.E., Haberland M., Reddy T., Cournapeau D., Burovski E., Peterson P., Weckesser W., Bright J., van der Walt S.J., Brett M., Wilson J., Millman K.J., Mayorov N., Nelson A.R.J., Jones E., Kern R., Larson E., Carey C.J., Polat I., Feng Yu, Moore E.W., VanderPlas J., Laxalde D., Perktold J., Cimrman R., Henriksen I., Quintero E.A., Harris C.R., Archibald A.M., Ribeiro A.H., Pedregosa F., van Mulbregt P., SciPy 1.0 Contributors. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nat. Methods, 2020, vol. 17, iss. 3, pp. 261–272.Wäldchen J., Mäder P. Plant Species Identification Using Computer Vision Techniques: A Systematic Literature Review // Archives of Computational Methods in Engineering. 2018. Vol. 25, iss. 2. P. 507–543.Wäldchen J., Mäder P. Plant Species Identification Using Computer Vision Techniques: A Systematic Literature Review. Archives of Computational Methods in Engineering, 2018, vol. 25, iss. 2, pp. 507–543.Weinstein B.G., Marconi S., Bohlman S., Zare A., White E. Individual tree-crown detection in rgb imagery using semi-supervised deep learning neural networks // Remote Sens (Basel). 2019. Vol. 11, iss. 11. Art. no. 1309.Weinstein B.G., Marconi S., Bohlman S., Zare A., White E. Individual tree-crown detection in rgb imagery using semi-supervised deep learning neural networks. Remote Sens (Basel), 2019, vol. 11, iss. 11, art. no. 1309.Weinstein B.G., Marconi S., Bohlman S., Zare A., White E. Cross-site learning in deep learning RGB tree crown detection // Ecol. Inform. 2020. Vol. 56. Art. no. 101061.Weinstein B.G., Marconi S., Bohlman S., Zare A., White E. Cross-site learning in deep learning RGB tree crown detection. Ecol. Inform., 2020, vol. 56, art. no. 101061.Weinstein B.G., Graves S.J., Marconi S., Singh A., Zare A., Stewart D., Bohlman S., White E. A benchmark dataset for canopy crown detection and delineation in co-registered airborne RGB, LiDAR and hyperspectral imagery from the National Ecological Observation Network // PLoS Comput. Biol. 2021. Vol. 17, iss. 7. Art. no. e1009180.Weinstein B.G., Graves S.J., Marconi S., Singh A., Zare A., Stewart D., Bohlman S., White E. A benchmark dataset for canopy crown detection and delineation in coregistered airborne RGB, LiDAR and hyperspectral imagery from the National Ecological Observation Network. PLoS Comput. Biol., 2021, vol. 17, iss. 7, art. no. e1009180.Zhong Lihui, Dai Zhengquan, Fang Panfei, Cao Yong, Wang Leiguang. A Review: Tree Species Classification Based on Remote Sensing Data and Classic Deep Learning-Based Methods // Forests. 2024. Vol. 15, iss. 5. Art. no. 852.Zhong Lihui, Dai Zhengquan, Fang Panfei, Cao Yong, Wang Leiguang. A Review: Tree Species Classification Based on Remote Sensing Data and Classic Deep Learning-Based Methods. Forests, 2024, vol. 15, iss. 5, art. no. 852.

The authors declare that there are no conflicts of interest present.