作者: Guijun Yang , Chunjiang Zhao , Qiang Liu , Wenjiang Huang , Jihua Wang
DOI: 10.1109/TGRS.2010.2071416
关键词:
摘要: This paper presents a new forest leaf area index (LAI) inversion method from multisource and multiangle data combined with radiative transfer model the strategy of -means clustering artificial neural network (ANN). Four scenes Landsat-5 Thematic Mapper (L5TM) Beijing-1 small satellite multispectral sensors (BJ1) images, acquired at different times, were selected to construct image in this study. Considering vertical distribution LAI both overstory understory, hybrid invertible reflectance (INFORM) was used support retrieval eliminate dependence understory vegetation. The simulated INFORM outputs, added random noise, first clustered by method, then trained ANN obtain for each group (cluster). Next, applied combinations retrieve LAI. Finally, validation inverted results Moderate Resolution Imaging Spectroradiometer product field measurements conducted. experimental indicate that accuracy can be improved through addition observation angle data, if quality is ensured. 30% compared average single after considering noise training data.