Font Size: a A A

Multi-scale Segmentation And Classification Of Geological Remote Sensing Imagery Based On Stepless Scale Technology

Posted on:2021-07-06Degree:MasterType:Thesis
Country:ChinaCandidate:Z X XiaoFull Text:PDF
GTID:2480306470488554Subject:Geology
Abstract/Summary:PDF Full Text Request
With the rapid development of aerospace remote sensing(RS)technology,the spatial,spectral,and temporal resolution of image has been continuously improved,providing rich texture,structure,and spatial information for feature recognition and classification.However,how to efficiently process and interpret high-resolution RS image is still one of the major difficulties in the field of remote sensing.On this background,object-oriented classification technology has begun to receive widespread attention,which is different from the previous pixel-based methods.It can avoid the “salt and pepper” in the classification results via dividing and merging local homogenous pixels as the basic classification unit.Undoubtedly,this method promotes the organic combination of the spatial and attribute characteristics of objects in classification using high-resolution RS imagery.Based on the continuously variable scale technology,we select an area to conduct objectoriented classification using WorldView-2 imagery.Several insights are given as follows:(1)Nearest neighbor,watershed and graph cut algorithms are used for image segmentation.The optimal segmentation scale is determined through multiple experiments.The graph cut algorithm has a comparative advantage in running time.According to the spectral and geometric information characteristics of different lithologies,a classification rule set is formulated,and K nearest neighbor and support vector machine methods are used for lithological classification.The accuracy evaluation shows that the segmentation classification results are more reliable.(2)The basic dykes have the smallest width in the study area,and the environment,terrain and weathering of the bedrock disturbs the extraction of it.Repeated experiments help us get the optimal extraction conditions: near N-S main direction?large aspect ratio and the small R,G,and B values.(3)The basic dykes,basic-ultrabasic rocks,intermediate-acid rocks and other rock units extracted in this study have important practical significance for the tectonic structure,mineralization and prospecting in the northern margin of Tarim.
Keywords/Search Tags:WorldView-2 imagery, Continuously variable scale, Multi-scale segmentation, Object-oriented classification, Baishan area of Ruoqiang county
PDF Full Text Request
Related items