Information Extraction And Pattern Analysis Of Rural Settlements In Hilly Areas Of Central Sichuan

Accurate poverty alleviation and rural revitalization are in full swing.Accurate,rapid and economical grasp of the size and spatial distribution of rural settlements is of great practical significance for the effective use of rural land,the rational planning of rural layout and the promotion of new rural construction.Taking Jinshi Town,Santai County,Mianyang City as an example,this paper studies the classification and extraction of rural settlements information based on three remote sensing images: GF-2,ZY-3 and Landsat-8,and analyses the spatial distribution pattern of rural settlements in hilly areas of central Sichuan.The main research contents and conclusions of this study are as follows:(1)Using e Cognition software to construct the extraction model of GF-2,ZY-3 and Landsat-8,and accurately obtain the distribution of the settlement in the study area in 2015.Firstly,the rate of change of local variance is used to select the optimal segmentation scale parameters,and establish a multi-scale segmentation system to complete image segmentation.Secondly,the spectral features,texture features,shape features and remote sensing derived data features of multiple level landcover types are discovered by analyzing selected samples,and used to construct the knowledge system of different landcover types classification,establish the model of extraction,and extract the landcover.Finally,the confusion extraction matrix and Kappa coefficient as well as other methods are used to evaluate the extraction accuracy.The overall accuracy of the GF-2,ZY-3,and Landsat-8 image extraction in the study area was 94.67%,88.68%,and 80.56%,respectively.(2)Comparative analysis of the effects of GF-2,ZY-3 and Landsat-8 images on the extraction of settlement information.The results show that the Landsat 8 and ZY-3 images are more advantageous for large-scale and low-precision extraction of large town,while the high-resolution GF-2 images are more advantageous for small and medium-sized villages and independent houses.(3)Using the spatial analysis function of GIS software,combined with the DEM,roads and rivers data of the study area to analyze the spatial pattern of rural settlements.There are 2,122 settlements in the study area,mainly small and medium-sized settlements.More than half of the settlements are under 500 square meters.The spatial distribution of settlements is mainly strip-like.More than 90% of the settlements are in the range of 0-500 m from the road.Most of the settlements are located in the first and second topographic index grade areas with better topographic conditions,gentle slope and smaller elevation.80.68% of the settlements are located in the area with an elevation less than 475 m,and the range of the settlements is the widest in the area with a slope of 2-6 degrees.And in this region,the number of settlements greatly decreases as the level of the terrain index increases.The junction of Quanyuan,Rongzhuang and Minglang Sancun is the most densely distributed area in the study area.(4)Explore the relationship between settlement area and population,cultivated land.The correlation between the settlement area and the population is significantly correlated at the 0.01 level,and the correlation coefficient is 0.897.Taking the rural settlement area as the independent variable and the population quantity as the dependent variable to establish the regression estimation model,the model's determination coefficient is 0.805.The model can be used to estimate the population of rural population based on the rural settlement area extracted by GF-2 image.