Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Chine → 2019 → Classification Study of Desert Steppe Species Based on UAV Hyperspectral Remote Sensing

Inner Mongolia Agricultural University (2019)

Classification Study of Desert Steppe Species Based on UAV Hyperspectral Remote Sensing


Titre : Classification Study of Desert Steppe Species Based on UAV Hyperspectral Remote Sensing

Auteur : 杨红艳;

Grade : Doctoral Dissertation 2019

Université : Inner Mongolia Agricultural University

Résumé partiel
In recent years,under the influence of climate change and human activities,grassland degradation is very serious,threatening ecological balance,causing grassland production reduction,sandstorms and other disasters.The change of dominant species in grassland ecosystems is a significant feature of grassland degradation,and some species have important implications in the process of grassland degradation.Therefore,rapid,non-destructive and large-area monitoring of grassland species is conducive to the correct judgment of grassland degradation degree,and is of great significance for grassland ecological environment management and grassland animal husbandry production.At present,due to the limitation of spatial resolution,grassland monitoring based on satellite remote sensing cannot reflect the structure of grassland species.Grassland monitoring based on field surveys is time-consuming and laborious,and it is difficult to meet regional monitoring needs.With the rapid development of unmanned aerial vehicle(UAV)and hyperspectral imaging technology,it provides a new method and technical basis for solving regional grassland species classification.In this study,the Inner Mongolia desert steppe was taken as the research object.The multi-rotor UAV was used as the remote sensing platform.The hyperspectral imager was installed on the UAV to collect the grassland hyperspectral remote sensing image during the grassland vegetation growth period.In view of the weak spectral difference of grassland species,it is difficult to directly use the spectrum to distinguish species.In this study,the spectral transformation method was used to increase the spectral gap of grassland species.On this basis,the vegetation index was established,and the classification threshold of the spectral transformation vegetation index was determined by the maximum inter-class variance method.The classification results show that the classification method based on spectral transformation vegetation index effectively extracts the spectral characteristics of grassland species,and the classification method is simple and feasible.In view of the data redundancy of spectral dimension inherent in hyperspectral imagery,considering the correlation of the hyperspectral bands,the effective information content of the bands and the separability of objects,a stepwise method was proposed to select the characteristic bands,which realizes the feature spectrum extraction and reduces spectral dimension.The hyperspectral images represented by the feature bands were used as the input data of the deep convolutional neural network(CNN)

Mots clés : Desert steppe; UAV; Hyperspectral remote sensing; Classification; CNN;

Présentation (CNKI)

Page publiée le 2 février 2021, mise à jour le 25 novembre 2021