Forest Canopy Cover Inversion Exploration Using Multi-Source Optical Data and Combined Methods

An accurate estimation of canopy cover can provide an important basis for forest ecological management by understanding the status and change patterns. The aim this paper is to investigate four methods random (RF), support vector regression (SVR), k-nearest neighbor (KNN), with fast iterative features selection (KNN-FIFS) modeling cover, evaluate three mainstream optical data sources—Landsat8 OLI, Sentinel-2A, Gaofen-1 (GF-1)—and types combined comparatively selecting optimal method. uses Daxinganling Ecological Station Genhe City, Inner Mongolia, as research area, based on multispectral remote sensing data, extracting spectral characteristics, textural terrain characteristics; Kauth–Thomas transform (K-T transform); color transformation characteristics (HIS). combination was selected using feature screening methods, namely stepwise regression, RF, KNN-FIFS, methods: SVR KNN, were carry out evaluation analysis regarding accuracy modeling: (1) In study, a variety introduced, variables different parameter preference then employed in modeling. Based inversion KNN-FIFS model achieves best accuracy: Landsat8 OLI R2 = 0.60, RMSE 0.11, RMSEr 14.64% model; Sentinel-2A 0.80, 0.08, 11.63% GF-1 0.55, 0.12, 15.04% federated 0.82, 10.40% (2) datasets have ability estimate superior under multi-source features; (3) KNN- FIFS established nonparametric model, its optimization method better than that For same result joint single data; thus, specific parameters, significantly improve efficiency from sources.