作者: John David Armston , Robert J Denham , Tim J Danaher , Peter F Scarth , Trevor N Moffiet
DOI: 10.1117/1.3216031
关键词:
摘要: The detection of long term trends in woody vegetation Queensland, Australia, from the Landsat-5 TM and Landsat-7 ETM+ sensors requires automated prediction overstorey foliage projective cover (FPC) a large volume Landsat imagery. This paper presents comparison parametric (Multiple Linear Regression, Generalized Models) machine learning (Random Forests, Support Vector Machines) regression models for predicting FPC Estimates were derived field measured stand basal area (RMSE 7.26%) calibration models. Independent estimates airborne LiDAR 5.34%) surveys validation model predictions. LiDAR-derived enabled bias variance predictions to be quantified regional areas. results showed all had similar errors < 10%), but less than at greater ~60% FPC. All 10% plant communities with high herbaceous or understorey this work indicate that use products data Queensland using any assumption senescent absent time image acquisition.