Remote Sensing Image Simulation Of Linear Array Satellite,Geometry Processing And Optimization

With the development of remote sensing science technology,the linear array images can efficiently excavate the earth surface resources,environment and other data and provide a lot of information.In the face of widely used remote array remote sensing images and related products,targeted to explore a geometric model,it not only can reduce the impact of linear satellite remote sensing system of internal and external factors caused by the changes of object position information of the image pixels corresponding to the place;but also can solve geometric model optimal solution surface optimization for different terrain complexity,when to use linear remote sensing image data fast and efficiently.The main research contents are as follows:1)Gauss surface topography function and satellite linear array imaging simulationSix kinds of mountainous terrain,which are generated by the Gauss surface terrain function,are used to obtain the truth-value of ground control information.VIEW function is used to simulate the attitude of the satellite sensor with different X,Z axis;the simulation of the 3D terrain uses vertical center projection imaging scanning.Considering complete each central projection strip,and generate a.PNG format picture,we can treat each strip as a two-dimensional array,and continuously to add other two-dimensional array into the first two-dimensional array,to achieve full control grid seamless image processing.2)Geometric processing of linear array image without elevation controlIn the actual operation of geometric correction,in order to eliminate the deformation difference of remote sensing image,we propose a non-elevation control vector data and raster image information by combining regional encryption iterative model registration method.For the raster image geometric correction,we use the feature attribute vector and raster data and typical geometric features,to get the same point;according to the global rough registration and local fine registration principle,gradual distribute control points,iterative registration,and use the least median squares(Least median of squares,LMedS)to assess the matching error,the LMedS method has good robustness to the matching error,can eliminate the gross error control points,thereby reduce the overall RMS error,and improve the accuracy of matching.In this paper,it is significant to eliminate the large deformation of the image and reduce the difficulty of geometric correction,and it is a useful supplement to the topographic correction method.3)Geometric processing of linear array image with elevation controlThe rational function model(Rational Function Model,RFM,generally referred to as RPC model,Rational Polynomial Coefficient)are studied,combined with geometric position of remote sensing image and the parameters of RPC model for linear array image solution method on elevation control information for geometric processing,in order to avoid the ill conditioned matrix and over parameterization the problems appeared in the process of solving the parameters of RPC model,it can use partial least squares method.Six kinds of terrain simulation form and SPOT5 image data are divided into five kinds of elevation stratification to discuss the application scope of different RPC models,and explore the optimal solution of the RPC model.Based on the prior knowledge of the optimal result,the error sources and the system error compensation in the process of data processing are analyzed in this paper.Based on the error processing,analysis of the stepwise regression,partial least squares,orthogonal distance regression,least squares spectrum correction and ridge estimation method to solve the parameter of RPC model,the accuracy of the RPC model is compared with the different algorithms to get the optimal RPC model parameters.4)Experiment and conclusionAccording to the algorithm and experiment,we consider the RPC model with 9 different kinds of special model and application environment,and combined with the MATLAB simulation of 3D terrain scanning imaging and SPOT5 image data for experiments,to study the solution and optimization for different terrain under the form of the RPC model.The test results show that the RPC model with the geometric correction accuracy than the denominator form with no form of denominator model is higher,In 9 kinds of RPC models,the third most widely applicability,ability to correct for the best image;For flat terrain image,a first-order RPC model can form image correction operation with satisfying accuracy;for complex terrain of large terrain simulation and SPOT linear array images,in the form of three order model RPC model forms,even two order without the denominator will be able to complete the image rectification.