Application of Image Recognition Methods to Determine Land Use Classes DOI Creative Commons

Julius Jancevičius,

Diana Kalibatienė

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(9), P. 4765 - 4765

Published: April 25, 2025

The increasing availability of satellite data and advances in machine learning (ML) have significantly enhanced land use image classification for environmental monitoring. However, the primary challenge using imagery lies presence cloud cover, variations resolution, seasonal changes, which impact accuracy reliability. This paper aims to improve assessment cover changes by proposing a hybrid ML, interpolation, vegetation indices-based approach. proposed approach was implemented random forest (RF) classifier, combined with interpolation indices, classify Sentinel-2 Baltic States. experimental results demonstrate that achieves an rate above 90%, effectively demonstrating its capacity distinguish between various types. We believe this study will inspire researchers practitioners further work towards applying ML algorithms offer valuable insights future tasks involving noise digitalization research.

Language: Английский

Forest Height and Volume Mapping in Northern Spain with Multi-Source Earth Observation Data: Method and Data Comparison DOI Open Access
Iyán Teijido-Murias, Oleg Antropov, Carlos A. López‐Sánchez

et al.

Forests, Journal Year: 2025, Volume and Issue: 16(4), P. 563 - 563

Published: March 24, 2025

Accurate forest monitoring is critical for achieving the objectives of European Green Deal. While national inventories provide consistent information on state forests, their temporal frequency inadequate fast-growing species with 15-year rotations when are conducted every 10 years. However, Earth observation (EO) satellite systems can be used to address this challenge. Remote sensing satellites enable continuous acquisition land cover data high (annually or shorter), at a spatial resolution 10-30 m per pixel. This study focused northern Spain, highly productive region. aimed improve models predicting variables in plantations Spain by integrating optical (Sentinel-2) and imaging radar (Sentinel-1, ALOS-2 PALSAR-2 TanDEM-X) datasets supported climatic terrain variables. Five popular machine learning algorithms were compared, namely kNN, LightGBM, Random Forest, MLR, XGBoost. The findings show an improvement R2 from 0.24 only Sentinel-2 MultiLinear Regression 0.49 XGboost multi-source EO data. It concluded that combination datasets, regardless model used, significantly enhances performance, TanDEM-X standing out remarkable ability valuable height volume, particularly complex such as Spain.

Language: Английский

Citations

0
Application of Image Recognition Methods to Determine Land Use Classes DOI Creative Commons

Julius Jancevičius,

Diana Kalibatienė

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(9), P. 4765 - 4765

Published: April 25, 2025

The increasing availability of satellite data and advances in machine learning (ML) have significantly enhanced land use image classification for environmental monitoring. However, the primary challenge using imagery lies presence cloud cover, variations resolution, seasonal changes, which impact accuracy reliability. This paper aims to improve assessment cover changes by proposing a hybrid ML, interpolation, vegetation indices-based approach. proposed approach was implemented random forest (RF) classifier, combined with interpolation indices, classify Sentinel-2 Baltic States. experimental results demonstrate that achieves an rate above 90%, effectively demonstrating its capacity distinguish between various types. We believe this study will inspire researchers practitioners further work towards applying ML algorithms offer valuable insights future tasks involving noise digitalization research.

Language: Английский

Citations

0