Study Of Land Cover Classification Based On Multi-scale Feature Fusion In Yinchuan Plain

With the development of remote sensing technology,high-resolution remote sensing as a new type of remote sensing has become an important means for people to obtain information on the ground,and it also plays an extremely important role in agriculture and industry.Urban land cover is one of the main fields of remote sensing application.Remote sensing technology provides a rich information source and implementation method.Image fusion technology is the prerequisite for realizing various applications.Image fusion is the process of integrating multiple images of the same target source into one image.The fused image can obtain more comprehensive and accurate information than a single image.Aiming at the characteristics of multispectral and panchromatic remote sensing images.In this essay,an improved non-subsampled shearlet transform(NSST)fusion algorithm is designed.The original multispectral image is transformed by IHS space,and then the non-subsampled shearlet transform is used to obtain the low frequency sub-band and the high frequency direction sub-band.In this essay,the weighted method of regional energy adaptive selection is proposed as the standard to construct low frequency sub-band method.The high frequency direction sub-band uses the spatial frequency and Laplacian energy and the characteristic as the input motivation of the pulse coupled neural network PCNN.And the relative output fire map is adopted to select the pixels in the sub-band images,Finally,the inversion and IHS inverse transform of the non-subsampled shearlet transform are used to obtain the fused image.Compared with the other fusion algorithm,the experimental results have obtained better fusion clarity effect no matter in subjective evaluation or in objective evaluation,and provide more accurate information for high-resolution images in land-use change analysis.Based on the fusion algorithm,this essay studied the change of land type of high-resolution remote sensing images in Jinfeng District of Yinchuan City in 2000,2006,2010,2013 and 2015,whose data sources are mainly from the ZY-1-02 C and Landsat images,The non-supervised classification of the five data was carried out on the platform of ENVI software,and the general classification map was obtained,and the results of landcover classification in Jinfeng area were analyzed statistically to obtain nearly 15 years of land cover change information.The results showed that the land cover in the study area was significantly changed during the 15 years,and the construction area was significantly increased from 2000 to 2006.The period of 2000-2006 and 2006-2010 was the rapid expansion period of Yinchuan City.In 2010-2015,with the mature development in the commercial real estate engineering and ecological construction projects,and the rapid development in public hall,Green Park and other construction make Jinfeng District urbanization gradually forming.The results show that the rapid development of Yinchuan city provides a reference for land production management and balance and stability in the region.