地理空间信息2025，Vol.23Issue(4)：71-74,4.DOI:10.3969/j.issn.1672-4623.2025.04.015
基于GNSS-R数据和机器学习算法探测森林火灾研究
Research on Forest Fire Detection Based on GNSS-R Data and Machine Learning Algorithm
摘要
Abstract
Based on GNSS-R and accumulated fire bum area data,we used convolutional neural network(CNN)and random forest(RF)algo-rithm to detect fire in the southeastern region of Brazil.The results show that the CNN method significantly outperforms the RF method in terms of congruence with actual fire areas,exhibiting fewer anomalies.Relative error in detecting fire areas is less than 6％for both methods,with the CNN method demonstrating 3.3％increase in accuracy and 17.4％enhancement in model fit over the RF algorithm.Error analysis indicates that the detection errors for both methods adhere to a normal distribution,suggesting an absence of significant systemic errors in the models.Spatial distribution analysis reveals that the fire areas detected by the CNN method align more closely with the actual conditions,in contrast to the RF al-gorithm which tends to overestimate the actual fire areas.This research offers a novel perspective in utilizing GNSS-R technology for fire detec-tion,holding significant implications for future fire prediction and prevention strategies.关键词
全球导航卫星系统反射技术/火灾/卷积神经网络/随机森林/相对误差Key words
GNSS-R/fire/CNN/RF/relative error分类
天文与地球科学引用本文复制引用
杨金虎..基于GNSS-R数据和机器学习算法探测森林火灾研究[J].地理空间信息,2025,23(4):71-74,4.基金项目
中煤科工集团重庆研究院有限公司自立重点项目(2023ZDYF15) （2023ZDYF15）
山西焦煤能源集团股份有限公司西山分公司自立项目(23140109202306G),中国煤炭科工集团有限公司科技创新创业资金专项国际合作项目(2020-2-GJHZ003). （23140109202306G）
