Feature extraction for urban vegetation stress identification using hyperspectral remote sensing

摘要

Urban vegetation is commonly in stress state, due to environmental contaminations, poor soil, water and other human activities and intervention. This article took the principal means by vegetation abundance extraction, studies based on Hyperion hyperspectral image of east Guangzhou urban district, went through the steps: hyperspectral image preprocess — feature extraction — SMACC pixel unmixing, to extract 6 kinds of vegetation abundance images. Forward further purified these vegetation abundance images by PPI iteration, and then extracted seven kinds of endmembers from vegetation health characteristics differences, to obtain the index image of different vegetation stress intensity levels. We through on-site investigation to check out the circumstances around stress site, to analyze the vegetation stress causes. Combined with vegetation spectrum analysis between vegetation abundance and stress. This investigation has more accuracy than vegetation indices calculation, provides accurate material for the urban green space investigation management.