Study On Cloud And Snow Discrimination In The Highlighted And Complex Underlying Surface

Satellite remote sensing technology has been widely used in geological disaster monitoring,dynamic monitoring of water systems,agricultural resource surveys,environmental protection and other fields.Among them,multispectral data has the characteristics of relatively high spatial resolution,temporal resolution,and rich spectral information.Therefore,it is the most widely used in the field of remote sensing.In recent years,the free opening of multi-spectral satellite data represented by Landsat 8 has provided a rich source of data for the application of multi-spectral satellite remote sensingCloud detection is an essential step in the application of remote sensing images.At present,there are still some problems that need to be solved in the process of cloud detection of remote sensing image.For example,most cloud detection methods have good detection effects in the thick cloud region,but for thin or mist-shaped clouds,there are often cases of missed detection.The spectrum of the highlighted ground surface features such as clouds,ice,snow,or deserts is highlighted in the visible light band.When the underlying surface has objects such as ice,snow,or deserts,which have similar spectral characteristics to the clouds,the cloud detection effect is poor and easy Multi-judgment occursTherefore,based on the above background,the research content of this paper is to separate the cloud and snow from the surface area of the bright and complex underlying surface,and research and construct three algorithms for cloud detection and cloud and snow separation.The main research results are as follows:1.The characteristic absorption and reflection characteristics of ground feature information of cloud and highlighted complex underlying surface in different bands,as well as the complementarity and redundancy between different ground feature spectra are analyzed It is concluded that band 4,band 6,band 7 and band 9 have the diagnostic characteristics of distinguishing cloud from snow and highlighted complex underlying surface2.Tested the application effect of the newly added cirrus cloud band in Landsat 8,focused on the advantages and disadvantages of its application in cloud detection,and summarized several major problems that still exist in this band3.Researched and built a cloud detection algorithm based on principal component analysis and fractal summation model.In the area of the underlying surface,the algorithm has an ideal detection effect,especially for point clouds which are difficult to be identified.The thin cloud layer on the edge of thick cloud,the effect of this algorithm is remarkable4.Researched and constructed a cloud and snow separation algorithm for the highlight underlay area.The algorithm can separate the cloud from the salt and pepper noise generated by the highlight underlay area such as snow and desert,and further divide the cloud into thin clouds.,medium clouds,thick clouds.Especially when there is a lot of snow on the underlying surface of the Landsat 8 image or contains point clouds with small shapes that are difficult to identify,the separation effect of the cloud and snow is ideal5.Researched and constructed a cloud-snow separation algorithm based on multi-dimensional information.This algorithm is ideal for separating cloud and snow,especially when the Landsat 8 image contains more complex underlying surface information,or contains contour boundaries.The cloud and snow separation effect of this algorithm is more significant when the cloud is fuzzy and presents cloud-like information that is difficult to identify.6.Researched the indicators suitable for quantitative evaluation of Landsat 8 image cloud and snow separation algorithm,and finally used accuracy rate,recall rate,precision rate,and F1 score in this paper to quantitatively judge the effect of cloud and snow separation.A data set with a wide variety of features,large north-south spans,diverse climates and terrains,and covering a large range is selected.The algorithm in this paper is compared with the artificial neural network and Fmask algorithm that have achieved good results in cloud detection,which proves the effectiveness of the algorithm in the separation of cloud and snow in the surface area of the highlighted and complex underlying surface.