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

Accueil du site → Master → Chine → 2009 → Utilization of Multispectral Images Segmentation Technology in Desertification Prevention and Cure

Tianjin University of Technology (2009)

Utilization of Multispectral Images Segmentation Technology in Desertification Prevention and Cure

Ma Jing Hua

Titre : Utilization of Multispectral Images Segmentation Technology in Desertification Prevention and Cure

Auteur : Ma Jing Hua

Grade : Master’s Theses 2009

Université : Tianjin University of Technology

Résumé
TM multi-band remote sensing image surface features and its classification has been puzzling questions in the field of remote sensing image processing . Due to the wide variety of surface features , and a variety of different feature changing with the passage of time and spatial distribution of imaging spectrometer as a multi-band surface reflectance data is also constantly changing . In addition, the imaging spectrometer data also exist isomer spectrum heterogeneity of the same spectrum of complex phenomena ; interference also be influenced by a variety of factors in the image data acquisition process . These unfavorable factors to the traditional classification methods exist to calculate excessive classification accuracy , generalization ability of poor shortcomings . The desert terrain multispectral image also has a significant addition to the general multi- spectral image has the characteristics of texture features . In this study, in the analysis of target characteristics and the general image classification method , the desert terrain remote sensing image data characteristics on the basis of , the proposed method of combining support vector machine and texture features , to overcome the defects of general image segmentation methods . Through experiments with of other multi Guangpu image segmentation method , it is found that the SVM method can achieve better segmentation results . Traditional methods such as Bayesian methods and neural network methods are based on the empirical risk minimization theory . The support vector machine is based on the theory of structural risk minimization . It is mainly used in which the environment of the lack of samples , and the ability to deal with the problem in many areas , such as model selection problems , linear programming problems , as well as strike a local minimum point . Taking into account the multi-spectral images of the desert terrain characteristics has the texture feature , texture feature and SVM combined , the results of the experiment show that this method has higher classification accuracy .

Mots clés : Support Vector Machine Multi-spectral image Machine Learning Texture features

Présentation (Dissertation Topics)

Page publiée le 22 mai 2013, mise à jour le 11 mars 2018