Ecological Systems and Devices Annotation << Back Classification Methods for Water Protection Zone Elements in Aerial Imagery for Ecological Assessment and Anthropogenic Impact Analysis Fomina E.S., Saenko A.A., Mironov A.S. The paper addresses issues related to identifying signs of violations in nature conservation and ecosystem states within water protection zones of rivers. It presents the results of a study conducted in the water protection zone of the Amur River in Khabarovsk district, Khabarovsk region, spanning the years 2024–2025. The study employs automated analysis of aerial photographs obtained via unmanned aerial vehicles (UAVs) using machine learning methods. Different machine learning techniques are discussed for classifying fragments of aerial photographs to detect various elements of the water protection zone, changes in which may indicate violations in nature conservation and ecosystem degradation. Both classical machine learning methods applicable to computer vision tasks (Naive Bayes, k-means clustering, Support Vector Machines, and Random Forest) and their ensembles, as well as deep learning models, are analyzed. Classifier performance is evaluated based on the percentage of correctly identified objects and the F-measure. A novel approach is introduced, combining a deep learning model, classical k-means clustering, and a multilayer perceptron into a pipeline. This approach enhances the classification quality of the water protection zone by 4.5 % and the detection of nature conservation violations by 8.2 % compared to conventional deep learning models. Keywords: ecological assessment, aerial photographs, classification, violations of nature management, machine learning methods. DOI: 10.25791/esip.10.2025.1549 Pp. 21-30.