Research On Spatial Allocation Method Of Sample Points Based On Remote Sensing Data

Frequent human activities aggravate land use and land cover changes.How to obtain a reliable remote sensing classification map is the most fundamental and key point.At the same time,the location distribution of sample points in the evaluation of remote sensing classification accuracy is an important factor in the evaluation results of remote sensing classification accuracy.In view of the problem that the sample points in the current remote sensing classification accuracy evaluation are usually difficult to achieve the uniform distribution of attribute characteristics and geographical location,this study takes Beijing-Tianjin-Hebei as the research area,and comprehensively utilizes Sentinel-2,Landsat 8,MODIS and various remote sensing data products.To carry out the research on the spatial distribution method of sample points based on remote sensing data and a layered and uniform allocation method of sample points for remote sensing classification accuracy evaluation is obtained.The research process and results of this article are as follows:(1)Firstly,spatial stratification based on multi-scale remote sensing data products is studied.Taking Beijing-Tianjin-Hebei as the research area,using 2017's10m resolution FROM-GLC10,30 m resolution NLCD,500 m resolution MCD12Q1 and other remote sensing data products after reclassification,considering the spatial heterogeneity and boundary area,A spatial layering method combining spatial heterogeneity and eight-neighbor algorithm is proposed.The research results show that the method considers the influence of spatial heterogeneity and boundary area,distinguishes the inner and outer boundaries of the type,and the final spatial layer number is 24 layers.(2)Secondly,the attribute characteristics of sample points and geographical location allocation methods is studied.Taking Beijing-Tianjin-Hebei region as the study area,five groups of data sets were selected by using different sizes of grids,and the total number of sample points in each group was 13503,6002,3368,2170 and1505,respectively.The attribute feature distribution of sample points was carried out according to the area ratio between each layer of the spatial stratification results;the optimization algorithm of the sampling point objective function combined with the space simulated annealing algorithm is constructed by the average shortest distance minimization criterion written in R language to complete the geographical location distribution of the sample points;at the same time,the geographical location distribution of the sample points of other allocation methods is completed.(3)Finally,the accuracy evaluation of space allocation methods for different sample points is studied.Take Beijing-Tianjin-Hebei as the research area,and set up comparative experiments of different sample point allocation methods.The research results show that overall accuracy,relative accuracy,root mean square error and standard deviation of the layered uniform distribution method in this study are67.272%?67.981%,99.281%?99.876%,0.284% and 0.302%,which are better than simple Random allocation method,space uniform allocation method and stratified random allocation method ensure the accuracy and reliability of remote sensing classification accuracy evaluation.The spatial allocation method of sample points based on remote sensing data studied in the thesis can be applied to the evaluation of remote sensing classification accuracy,the detailed investigation of soil pollution,the quality detection of agricultural land grade,the authenticity test and so on.It is helpful to improve the quality of sample point data,reduce data redundancy,and ensure the accuracy and reliability of remote sensing classification accuracy evaluation.At the same time,it provides a new technical method for the authenticity test field.