Research On On-orbit Geometric Calibration And Automatic Processing Technology Of Remote Sensing Sensors

Optical remote sensing satellites can acquire stereoscopic images on a global scale.Using strict imaging models and precise geometric parameters of sensors,it is possible to realize many important military applications such as reconnaissance and precise positioning of military targets,global mapping without control points,and precise guidance of weapons.The on-orbit geometric calibration of remote sensing sensors is the key technology for accurate positioning of sensors.At present,the implementation of on-orbit geometric calibration tasks usually requires a certain amount of manual operations,the degree of automatic processing capability of geometric calibration is low,and the accuracy of geometric calibration needs to be further improved.With the increasing number of remote sensing satellites operating in orbit and the increasing geometric resolution of satellite imagery,the automatic processing and higher accuracy of on-orbit geometric calibration has become an urgent need.This dissertation focuses on the key techniques of automatic processing of on-orbit geometric calibration for remote sensing satellite sensors.The digital calibration field and the reflection point source target are taken as research objects of the dissertation.The main research contents contain on-orbit geometric calibration automatic processing based on digital calibration field,automatic and intelligent identification and measurement of point source target image,control network structure optimization and the improvement of geometric calibration accuracy.The practical method of the on-orbit geometric calibration of remote sensing satellites is explored,which provides the basis for the automated on-orbit geometric calibration of domestic high-resolution remote sensing satellites.The main contents and innovations of the dissertation are listed as follows:1.The TH-1 satellite is used as an example to analyze the imaging characteristics of the linear array sensor.The time,orbit and attitude data acquisition principles of the sensors are introduced.The rigorous imaging model and geometric calibration model of satellite sensors are derived.Combined with the actual satellite attitude and orbit data,the problems about parameters variation law and parameter correlation in the geometric calibration model are analyzed and the applicable range of different geometric calibration strategy is explained.Then different types of geometric calibration facilities such as traditional artificial targets,natural and artificial features,digital calibration fields,reflective point source targets are discussed.The advantages and disadvantages of different geometric calibration facilities are analyzed,and the advantages and disadvantages of different types of targets for on-orbit geometric calibration automatic processing are discussed.2.Considering the difficulty of high-precision DOM and low-resolution satellite image registration,a multi-level image automatic geometric registration algorithm suitable is proposed,which supports geometric calibration automatic processing based on digital calibration field.Registration between remote sensing images with large differences is studied to provide technical support for automatic geometric calibration based on digital calibration field.A multi-stage image registration method combining the feature based and area based registration method is proposed.The registration is completed in three stages,avoiding the direct use of the original image,thus reducing the difficulty of automatic geometric registration.The registration is realized with the proposed method for a variety of remote sensing images with resolution difference ranging from 1 to 50 times.The registration accuracy is verified to be sub-pixel level.Then,the dense control points are acquired automatically based on this registration method,and the geometric calibration of the satellite sensor is completed.Finally,the validity of the calibration results was verified using multiple sets of data from the same sensor in different phases and regions.3.Aiming at the low degree of automation of traditional control point selection,an automatic recognition method for point source target image control points based on image features is proposed.The imaging and application principles of point source target are introduced,which includes layout,automatic star searching and optical path alignment.Then the characteristics of point source target images are analyzed.Based on the image characteristics of point source target,an automatic recognition method for point source target images is proposed.The prior knowledge and image feature parameters are used to accurately identify the point source target image accurately.Automatic identification method of point source target image control point requires a priori information,and the identification speed is low.Aiming at this problem,a fast intelligent recognition algorithm for point source target images based on the simulated dataset is proposed.The neural network model trained by the point source target image simulated dataset can accurately and quickly identify the real point source target image without relying on any prior information.The centroid location of the identified point source target image is calculated by three methods,and the result verifies that the point source target image has high relative measurement precision.4.The influence of control point pixel error,quantity of control points,and control network structure on geometric calibration and positioning accuracy is studied when different targets are used for geometric calibration.Firstly,based on the principle of image degradation,the simulated point source target images are generated and their pixel measurement accuracy is calculated.The satellite and airborne simulation data used to verify the geometric calibration accuracy are generated by using the actual data.Geometric calibration is applied on the satellite simulated data by using control points with different pixel errors.The calibration parameters are solved and the positioning accuracy using the solved calibration parameters is verified by checkpoints.It is found that the traditional artificial targets have a great influence on the external and internal calibration parameters.The calibration and positioning accuracy of check points calculated with calibration parametres using traditional targets is worse than that using the point source targets.Experiments have also shown that a small number of point source targets can achieve the same high calibration accuracy as denser traditional control points can.The experiments of airborne simulation data also verifies that the geometric calibration using the point source target can obtain more accurate calibration parameters and positioning results.