Study On Methods Of Extracting Geometrical Features From Remote Sensing Images With High Spatial Resolution

The weather, climate and global changes are not only the scientific but also environmentalproblems. With the rapid developments of spatial information science and technology, thesatellite imagery with high resolution have generated dramatically and formed satellite groupswith different resolutions. But the current status is that the methods of extracting informationfrom the great numbers of remotely sensed data cannot meet the actual requirement, andespecially the information use of the high spatial resolution remote sensing data requiresextracting the geometrical and structural information from images explicitly and effectively. Inaddition, the researches of land surface process models demand accurate parameters of therugged terrains. The dynamic updating of GIS data and automatic digital measurement also needto extract the geometrical features information precisely. Therefore, study on the methods ofextracting geometrical features from remote sensing images with high resolution will have greatsignificance and application values.Using mathematical morphology methods, in this paper, the methods of extractinggeometrical features from the remote sensing images with a high spatial resolution were mainlyintroduced. The purposed contents include the denoising algorithms with mathematicalmorphology filter algorithms, the remotely sensed image edge detection algorithms and theelimination of the image edge short glitch using mathematical morphology methods. Finally, theextraction model of geometrical features for remote sensing image was created. By tracing theimage, edge information can be achieved, then raster data and vector data also be generated.Based on the extraction model of geometrical features, the geometrical features of objects can beextracted. In this study, the processing of remote sensing images based on the mathematicalmorphology needs to not only design the algorithms, but also choose the shape and scale of thestructure element. Therefore, we design the multi-levels weighted filter algorithm usingomni-directional structure elements and multi-scale, omni-directional mathematical morphologyedge detection algorithm, which select the structure elements with different scales and directions,and successfully resolve the contradictions between noise suppression and fine edge extraction ofthe high spatial resolution remote sensing images, and own the perfect antinoise capability.The extracted geometrical features information can be used in object recognition andclassification of remote sensing image with its spectrums, texture, and statistics features. And theextracted information can be used in the fields of GIS, photographic surveying, computer visionand the industries of meteorology, agriculture and forest, ocean, hydropower, Land resources, andenvironmental protection.