Unequal Probability Sampling Method Based On Sub-compartments Objects And System Integration

Forest,as an important part of terrestrial ecosystem,is the material basis for the survival and development of human society.Forest stock volume is an important index to understand forest resources,so it is of great significance to estimate forest stock volume accurately.Taking Dongtai forest farm in Jiangsu Province as an example,this paper compares several sampling methods of forest stock volume estimation,and develops an unequal sampling system based on sub-compartments,so as to improve the informatization of sampling design,sampling survey and sampling estimation and improve the sampling efficiency.The main research contents and results are as follows:1.Exploring a new method of stratification cofactor acquisition based on deep learning modeland providing stratification cofactor of dominant tree species with high accuracy for forest stock volume estimation in the study area.Comprehensive utilization of this method is based on the deep learning model,the early stage of the survey data in the study area,2016unmanned aerial vehicle(UAV)high image data,supplementary investigation data in 2016,which is through a large advantage tree image feature extracting convolution neural network to realize information extraction and recognition of tree type,and get a small sample of stratification cofactor in the study area.The accuracy of the classification results of 4 species types was evaluated,and the area statistics of the fine identification results of the main dominant species were carried out,and the area statistics of each species in the supplementary survey data in 2016 were compared and analyzed.The results show that the proposed method based on the deep learning model for stratification cofactor acquisition has an overall accuracy of 91.18%and a Kappa coefficient of 0.88.The relative precision of Populus euramevicana,Metasequoia glyptostroboides,Ginkgo biloba,Bambusoideae and other forest land area components is 96.77%,84.28%,87.04%and 90.48%,respectively,and the relative average precision is 89.65%,which can meet the requirements of forestry production.2.Getting sampling accuracy of forest stock volume in the study area by several methods in order to make a comparison.This paper adopts five sampling methods:sub-compartments simple random sampling,sub-compartments stratified sampling,sub-compartments PPS(Probability Proportionate to Size Sampling)sampling,sub-compartments?PS sampling and sub-compartments stratified product?PS sampling.The stratification cofactor adopts the results of fine classification of tree species.The optimal sampling method was selected to estimate the total forest stock volume in the study area after comparing the sampling precision of the five sampling methods.The results show that the stratified?PS sampling method based on the stratified auxiliary information of sub-compartments has better performance.In the same sample size,the sub-compartments stratified?PS sampling is higher than simple random sampling of sub-compartments,sub-compartments stratified sampling,sub-compartments PPS sampling.Sub-compartments stratified?PS sampling precision is high,its sampling precision is up to 91.52%,the study area forest stock volume is estimated to be 211,992.03m~3.The low accuracy of PPS sampling is mainly due to the fact that PPS sampling is a sample method with replacement,which may lead to the repeated sampling of some very large forests and sub-compartments,affecting the estimation accuracy.The accuracy of several methods of equal probability sampling is low mainly because the advantage of unequal probability sampling is high efficiency.The?PS sampling method has the characteristics of low cost and high efficiency.It has a good application prospect in forest resources investigation.3.Developing the unequal probability sampling system based on sub-compartments objects.Server is based on PostgreSQL/PostGIS database technology.The system uses C#language in Visual Stodio2010 platform to complete the spatial data database application and interface design,and employ the Web Services technology to achieve spatial database access and application of the network interface;On the client side,the spatial data cache structure class is designed and implemented based on C#language,and the interface for caching and reading and writing of spatial data is realized.An application with Shapefile data import,export and query export function is designed through example analysis.We have successfully implemented an unequal sampling system for Android mobile devices.The main functions of the system including:(1)basic map operations:Shapefile reading and display,tile data loading and display,and basic map interactive operations,such as scaling,translation,positioning and search,etc.(2)layout and reset of sample sites:the functions of selecting sample sites,returning distribution point interface and marking investigated sample sites are realized;(3)sample data collection:including spatial data,sample ground survey data and Angle gauge survey data;(4)statistical analysis:to estimate the forest stock volume of the study area through the?PS sampling algorithm,and obtain the sampling precision,area statistics and forest stock volume estimation;(5)data access:the mobile terminal can query the current data and download the historical data through the server.