Quantitative Monitoring Of Soil Erosion In Yuelu District Of Changsha Based On Subpixel Mapping Technology

Quantitative monitoring of soil erosion is of great significance for the construction of national ecological civilization and the construction of a beautiful China.In recent years,remote sensing based soil erosion monitoring models have become an important technical means in quantitative monitoring of soil erosion,and have been widely applied.However,in the context of complex urban environments,there are a large number of mixed pixels in urban remote sensing images,and traditional pixel level soil erosion models are difficult to support high-precision quantitative monitoring of soil erosion.How to effectively utilize the abundance information of each end element and the spatial information of sub pixels in mixed pixels to improve the accuracy of quantitative monitoring of soil erosion is a focus and difficulty in existing research.Therefore,this thesis uses sub-pixel mapping technology,combined with the modified general soil erosion model,to conduct quantitative monitoring and spatial analysis of soil erosion in Yuelu District of Changsha City in 2020.The main research content of this article is as follows:(1)A remote sensing image subpixel mapping method based on super-resolution reconstruction convolutional neural network is proposed using super-resolution reconstruction technology to address the issue of excessive discrete noise in traditional subpixel mapping results.This method uses the Enhanced Deep Super Resolution Network(EDSR)to reconstruct the low resolution abundance map generated by the decomposition of mixed pixels through the transfer learning strategy.The experimental results show that the proposed method has higher accuracy compared to traditional SPSAM and RBF methods,and can effectively suppress the problem of high dependence of traditional subpixel mapping methods on mixed pixel decomposition results,which leads to more discrete noise in subpixel mapping results.(2)Due to the similarity of spectral features between shadows and water bodies,traditional subpixel mapping methods often have the problem of mistakenly identifying shaded areas as water bodies.To address this issue,a subpixel mapping method for urban land use that takes into account the influence of shadows is proposed.This method proposes an abundance optimization strategy to optimize the mixed pixel decomposition error caused by mountain shadows and high-rise building shadows.The experimental results show that this method can effectively overcome the influence of shadow errors in subpixel mapping,effectively reducing the phenomenon of shadow areas being mistakenly identified as water bodies,and further improving the accuracy of subpixel mapping compared to traditional methods.(3)In response to the problem that traditional soil erosion models are mostly based on pixel level and are difficult to support high-precision quantitative monitoring of soil erosion in complex urban environments,this thesis applies a sub pixel scale soil erosion model,Sub RUSLE.This model obtains sub pixel scale land use information through sub pixel mapping technology,establishes sub pixel scale soil and water conservation measures factor Sub-P,and obtains sub pixel scale soil and water loss model,improving model monitoring accuracy.Taking Yuelu District of Changsha City as the research area,an improved Sub RUSLE model was used to quantitatively monitor soil erosion in the study area in 2020.The research results indicate that the soil and water loss in the Yuelu District in 2020 was mainly characterized by low-level soil and water loss such as slight erosion and mild erosion,and the situation of soil and water loss was well controlled.At the same time,areas with severe soil erosion are mostly distributed in rainfall.