Study On Geo-Info Tupu Of Loess Landform Landscape Based On Remote Sensing Texture

Remote sensing image is reflection of landscape.On the Loess Plateau,morphology of "numerous rivers" was formed by loess under the internal and external forces.Under the constraint of this morphology,vegetation,soil,water,climate,human activity,and other geographic factors restrict and interact with each other and produce complex and diverse landscapes.Remote sensing technology enables these landscapes to be directly reflected by morphological characteristics of images.However,there is yet no proper solution for quantitatively expressing the landscape.This is mainly due to the lack of indicators that can describe the landscape at large scale and local scale simultaneously.Texture characteristics of remote sensing image describe the macroscopic characteristics as well as the localized characteristics.This paper employs the texture characteristics of remote sensing image to quantitatively express the morphological landscape.It starts with analysison the scale characteristics of loess landscape and scale effect of image texture.Related parameters are then optimized,which are employed to build texture characteristic graphics(Tupu)based on geographic information Tupu theory and methods.This Tupu is then applied to several areas such as the identification of morphological landscape type,the delineation of morphological zones,and the spatio-temporal variation of landscape.Contents and results of this paper include:(1)Study on the scale effect of loess morphological landscape and texture characteristics of remote sensing imagery.Wavelet analysis and semi-variance function analysis methods are used to study on the scale issue of loess morphological landscape.The results show that the texture characteristics of the different types of loess landforms are different.For the nine types of loess landforms studied in this paper,the appropriate scale is between 120 and 1920 meters.The variance characteristics of the texture index extracted by different texture models under different scale changes were evaluated by variance analysis according to the principle of intra-class homogeneity and inter-class heterogeneity.The results showed that spatial resolution of 15 m and analysis window size of 1201 × 1201 are appropriate scale parameters for studying the landscape features of loess landforms.The gray level co-occurrence matrix is the optimal texture model to extract the texture index.Image-based gray level co-occurrence matrix of second order,correlation between angular moment,LBP-based gray level co-occurrence matrix of second-order angle moment scale,contrast,and variance square sum are good indicators for expression of intra-class homogeneity and inter-class heterogeneity.(2)Construct remote sensing texture characteristic graphics(Tupu)of a typical landscape in the loess plateau of Northern Shaanxi.According to the geographic information Tupu theory and methods,remote sensing texture Tupu was constructed for a typical landscape in the loess plateau of Northern Shaanxi.Based on the seasonal variation of remote sensing texture Tupu,seasonal rhythm maps were generated.After careful comparative analysis,autumn images were selected for Tupu construction.(3)Explore the potential applications of Tupu in northern Shaanxi loess plateau landscape studies.After establishing a Tupu for typical landscape in the loess plateau of Northern Shaanxi,application potential of Tupu in the following four aspects of landscape studies were tested.a).Automatic recognition of landscape.Remote sensing texture recognition based on BP neural network landscape shows that remote sensing texture Tupu can help identify the type of landscape with an average accuracy of up to 90%.b).Delineation of landscape.According to different types of boundary,Tupu can help determine a clear boundary or transition zone on the margin of the different landscapes.c).Analysis of landscape spatial heterogeneity.Research area at the junction of Zizhou,Mizhi and Suide County exhibits more detailed spatial variation than current landscape delineation.This variation was found to be connected with terrain factors.d).Temporal analysis of morphological landscape.Using remote sensing image of 2000 and 2016,texture Tupu analysis was conducted in Yan'an city and a 123 km2 are around it.It was found that remote sensing texture Tupu reflects the landscape change induced by the construction of Yan'an new district,as well as the recovery effect of vegetation after the 1999 return farmland to forest policy was implemented in Yan'an.All these explorations reveal that the Tupu is effective to express macro landscape,to reveal the environmental effect caused by land cover change,and to evaluate the effect of return farmland to forest policy.It provides a useful support to the management,planning and governing of the eco-environment of Loess Plateau.In conclusion,this paper explores the feasibility of using remote sensing texture analysis for quantitatively expressing Loess Plateau morphological landscape.It starts with the understanding of corresponding scale effect of landforms landscape and remote sensing texture features.Multi-scale analysis method has been proposed under the basis of recognition of scale features and scale effects of landscape and remote sensing texture.Through quantitative analysis of texture features,methods and parameters for landscape expression are optimized.Based on geographic information Tupu theory and methods,the construction of remote sensing texture Tupu for typical landscape in the Loess Plateau of Northern Shaanxi provides a great support to further explore the landscape spatial pattern and variance.The demonstrated applications prove the potential of remote sensing texture Tupu in geography-related areas,and also its affectivity in providing reference to deeper geological applications.