Study On Construction Of Interpretation Samples And Feature Knowledge Base Of ZY-3 Satellite Image

High spatial resolution remote sensing technology has become the hotspot and popular technology in remote sensing field, and has been used in many fields. This paper uses ZY-3 satellite images as the research object, to study and discuss the general idea and basic strategy of construction of interpretation samples and feature knowledge base of ZY-3 satellite image; it uses methods including count, analysis and summery to get all kinds of feature information of some common target type such as water, vegetation, road and building area. Meanwhile, sample object, eigenwalues information etc are stored in the database. That realize search and call automatically, updates of samples set, enlarges and constraint call of sample characteristics knowledge.This paper makes the following concrete work:(1) It analyzes the existing land classification system, develops interpretation samples surface features classification system that suitable for domestic high resolution remote sensing images refer to characteristics of 2.1 meters resolution remote sensing images of ZY-3 and makes the sampling plan at the same time.(2) It sorts out 50 characteristics including spectrum, geometry, texture,index and etc., then quantitatively express and store them. The distribution regularity of eigenvalues of stair land type such as water, geometry, texture and index has been analyzed, then screening feature combination to build a feature vector to be the available assessment method.(3) Depends on samples and data type that stored in characteristics knowledge base such as characteristics of samples, vectors, images, eigenvalues and feature algorithm, this paper choices the spatial database solution of database platform Oracle and spatial database engine Arc SDE to build ZY-3 satellite image interpretation samples and feature knowledge base. Samples and feature knowledge base are developed based on C# and Arc GIS Engine to realize calling and maintaining database.(4) Taking the range area about 347 square kilometers in Binzhou, Shandong Province as test area, this paper collects 463 interpretation samples of water, vegetation, road and building area from image of ZY-3 to test the functions of interpretation samples and feature knowledge base. In addition, the paper also evaluates the availability of samples and fetching information depending on images at same area and latter phase.