Research On Geometric Correction Methods Of Jilin-1 Satellite Image

In recent decades,space remote sensing has developed rapidly at home and abroad,and the ground resolution,spectral resolution and temporal resolution of remote sensing images have been significantly improved.The remote sensing images with high ground resolution and high spatial positioning accuracy can replace the traditional aerial photogrammetry of UAV or the manual way of mining points to update topographic maps.The remote sensing images with high ground resolution characteristics can also provide realistic possibilities for human to carry out some fine ground object extraction and land change monitoring.The influence of the factors such as the internal and external errors of satellite sensors and the observation errors of the earth leads to the distortion of the image of optical remote sensing satellite,which results in the inadequacy of the positioning accuracy of image pixels to the ground,and can not meet the needs of scientific research and production such as space-time alignment of multisource data.How to improve the positioning accuracy of image pixels has very important practical significance.Satellite image vendors can improve the hardware performance of satellite sensors combined with image control information to a certain extent to improve the positioning accuracy of the original image without control points.At present,the positioning accuracy of domestic optical remote sensing images is still low without control points.The key to improve the positioning accuracy is to study the appropriate geometric correction processing method of optical remote sensing images.In view of the low accuracy of optical remote sensing image pixels in ground positioning in China,the factors affecting the accuracy of image geometric correction,such as geometric correction model,control point and digital elevation model(DEM),are studied in depth in this paper.In this paper,the geometric correction of optical satellite A panchromatic remote sensing image of Jilin-1 is studied by using different order polynomial model,local triangular network model and RPC rational function model combined with different number of control points and different resolution digital elevation model(DEM)to determine the reasonable geometric correction method of image.Compared with the original product image,the accuracy of image correction has been greatly improved.This paper mainly achieves the following results:(1)The positioning accuracy of Jilin-1 optical satellite A panchromatic image is slightly poor.After 12 uniformly distributed checkpoints statistics,the positioning accuracy of the original image without control points is about 80 m.(2)When the number of control points reaches a certain number,the accuracy of image geometric correction will not be improved with the increase of the number of control points.The experiment uses 0 to 15 control points to carry out geometric correction experiments on the panchromatic image of Jilin-1 optical star A.After 12 checkpoints statistical analysis,it is concluded that the correction accuracy of each correction model is no longer significantly improved or even slightly decreased after a certain number of control points have been reached.When the number of control points reaches 4,the image correction can achieve ideal accuracy results,and the RPC rational function model calibration accuracy reaches 1.810 m.(3)Under the same number of control points,the accuracy of RPC rational function model is higher than that of polynomial model and local triangulation model.The local triangular network model only corrects the image in the triangular network composed of control points.Geometric correction of the panchromatic image of Jilin-1 optical star A using RPC rational function model is the first choice to improve the accuracy of image pixel localization.(4)When RPC rational function model is used for geometric correction of the panchromatic image of Jilin-1 optical satellite A,the correction accuracy of digital elevation model using SRTM1 and SRTM3 is approximately the same under the same control point condition.When the number of control points reaches more than four,the accuracy of the corrected image can reach about 1.8m.