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

Accueil du site → Master → Chine → 2017 → Automatic Extraction of Desertification Information Based on GF-2 Spectral Features

Central South University Forestry and Technology (2017)

Automatic Extraction of Desertification Information Based on GF-2 Spectral Features

佘宇晨;

Titre : Automatic Extraction of Desertification Information Based on GF-2 Spectral Features

Auteur : 佘宇晨;

Grade : Master’s Theses 2017

Université : Central South University Forestry and Technology

Résumé
Stony desertification is an extreme form of land degradation and the most dire ct consequence of land resources lose.Besides,it will also lead lack of water,soil and soil fertility.Stony desertification,desertification,soil and water loss n ow becomes the three major ecological disasters in China.This research object is Guanshanhu,Guiyang city,Guizhou province.In the ba sis of remote sensing image in Guanshanhu(2006),by analysis stony desertific ation class data,second investigation and collateral data of stony desertification in Guizhou province and the survey data,finally a band algorithm that can dis tinguish between stony desertification land and non-desertification land.By use this method,not only made the distribution of rocky desertification,but also gr ading and spatial distribution analysis of Guanshanhu.By analyzing the monitor ing results of Guanshanhu,we find the causes and dynamics of desertification(By comparative analysis of the first and second monitoring results).Research results can provide scientific references for monitoring and control of rocky de sertification.The main research results of this paper include:1)According to analysis of GF-2 image gray scale,OIF index analysis and co mbination entropy,the combination of band 2、3、4 contains rich amount of i nformation,which is the best combination band in Guanshanhu informationextra ction.2)By the third rocky desertification monitoring standards of Guizhou province an d the properties of GF-2 image,finally determined the rocky desertification cla ssification system,which is divide into 5 parts:severe desertification,moderate desertification,slight desertification,potential desertification and non-rocky deser tification.3)Using maximum likelihood method to classify GF-2 image,it can effectively id entify the stony desertification land,non-desertification land,potential stony des ertification land.Classification of roads and building land boundaries is clear,a nd the true color images fits high plaques and gets well effect in stony deserti fication classification.In addition,in using object-oriented information extractio n to classify stony desertification information,the extraction and analysis result shows that segmentation scale =47,merging scale =95 are the best parameters in Guanshanhu.4)After accuracy evaluation,the final classification accuracy of maximum likeliho od method(after band math)is 83.2309%and 0.6752 of Kappa,is higher tha n maximum likelihood method(no band math)is 77.8021%and 0.6147 of Ka ppa,and also higher than object-oriented information extraction which final clas sification accuracy is 74.262%and 0.6233 of Kappa.After band math,the fina 1 classification accuracy improve 8.8689 percent.5)Completed rocky desertification information extraction process in using IDL,w hich is based on GF-2 spectral features.The operation of the module can quic kly and accurately extract the information of rocky desertification,and realize t he effective docking with the ENVI,which greatly raise the process of extracti ng the information of rocky desertification.This module includes two parts:the band operation module and the maximum likelihood module.6)The Guanshanghu stony desertification status and degree distribution map was been made.Comprehensive analysis of Guanshanhu desertification monitoring ar ea to such distribution,desertification and potential desertification of land and i ntensity distribution.In entire area of Guanshanhu,rocky desertification area ar e 3904.6 hectares.Severe desertification of 404.3 hectares,moderate desertificat ion of 2148.5 hectares,slight desertification of 1351.8 hectares,potential deserti fication of 9462.6 hectares,non-rocky desertification of 8156.5 hectares

Mots clés : Remote Sensing; Spectral Characteristics; Modular; Rocky Des ertification; Guanshanhu;

Présentation (CNKI)

Page publiée le 20 février 2018