| High-temporal and high-resolution thermal infrared remote sensing images are important resources for the researches of fine characterization of human traces,inversion of Earth surface features,resource exploration,and Marine ecological monitoring.The small CASEarth satellite is the first satellite supported by China's "Big Earth Data Science Engineering Project".The thermal infrared imager(TIRI),being one of the primary payloads in CASEarth,can obtain the thermal infrared radiation of the Earth in three spectral bands with a width of 300 Km and a resolution of 30 m.Generally,there are two types of data acquisition methods for remote sensing cameras with the linear detectors including the push-broom with a long linear array and the whisk-broom with a short linear array.However,affected by the engineering boundary constraint conditions such as structure size,weight,and power consumption of the satellite,the imaging method of push-broom with a long linear array is difficult to meet the requirements of short-term,wide-range and high resolution.Accordingly,the whisk-broom imaging with a long linear array is an effective method for solving this contradiction.However,because of the shortening of integration time and the increase of geometric positioning model parameters,the high-sensitivity and high-precision geometric positioning has become more difficult inevitably.In view of the CASEarth satellite TIRI which has the biggest ratio of imaging width and resolution in China,this paper focused on the research of high-precision geometric positioning method for the three spectrums paralleling whisk-broom camera with a multi-module splicing detector of 2000 pixels,established the rigorous geometric positioning model(RGPM)for the long linear whisk-broom TIRI,put forward the accurate measuring and calculating method of the imaging model for TIRI in laboratory and validated the feasibility of on-orbit calibration,overcame the difficulties in ground control points(GCPs)extraction affecting the calibration accuracy of TIRI,achieved the on-orbit image simulation based on the light trace tracking,DOM and DEM reference data for wide-range and high-resolution TIRI,and finally,the effectiveness of the RGPM and the proposed geometric calibration method are verified with the simulation data.Here,the main research contents and the corresponding innovations of this paper are summarized as follows:1.Based on orbit parameters of CASEarth satellite and the structures and imaging characteristics of the wide-range and high-resolution TIRI,we elaborately define the relative coordinates system and the corresponding transformations of the interior and exterior orientation model of TIRI,build the RGPM of TIRI,and analyze the influences of different model parameters on the positioning results and roles in model calibrating.And finally,according to the RGPM,the influences of various error sources on the geometric positioning accuracy of the image is systematically discussed,which lays a theoretical foundation for imaging simulation and geometric processing of the wide-range and high-resolution TIRI.2.Affected by the mechanics during launching process and the changes of in-orbit temperature field,the geometric positioning parameters including the installation matrix of scanning mirror and the interior parameters of TIRI need to be re-calibrated in orbit.This paper analyses the influences of different errors comprising the installation and position measurement results of scan mirror,displacement of principle point and distance,rotation and translation of the multi-module splicing detectors,and the tilt of focal plane,builds the auto self-calibration model for the whisk-broom TIRI with a multi-module splicing long linear detectors,proposes the parameters calculating method based on the least-squares theory,and implements a calibration accuracy less than 0.3 pixels based on the laboratory test data,which demonstrates the effectiveness of the imaging model calculation based on the GCPs and proposed model on-orbit.3.In terms of the difficulties in obtaining ground control information caused by the low spatial resolution,poor contrast,large difference of intensity mapping,as well as the indistinct high-dimensional features of thermal images,an accurate geometric texture-based GCPs extraction approach is proposed.This method makes full use of the plenty of geometric textural features of the remote sensing images,and employs Moravec algorithm,Sobel operator,adaptive filtering and morphological operations to extract the significant local textural patterns.And a geometric textural descriptor based on the Log-Polar transform is constructed for the acquired textural pattern map,which effectively avoids the heavy dependence of traditional GCPs extraction algorithms on the feature points and their surrounding gradient information.Meanwhile,for the existing remaining outliers from the conventional similarity matching,this paper proposes the double constrained matching rules comprising the matching bits and differences to eliminate the outliers,and realizes the matching extremal optimization during the cyclic shift of descriptor,which,greatly,reduces the effects of the mismatches on the precision of GCPs database.4.For the lack of on-orbit images of wide-range and high-resolution TIRI,according to the orbital parameters and RGPM of TIRI,reference image and DEM data,an imaging simulation method based on light trace tracking for the whisk-broom camera with a long linear detector is proposed to simulate the on-orbit imaging situation at any time and any positions.According to the RGPM of TIRI,the geometric calibration model(GCM)based on the "generalized" correction matrix is constructed.And the exterior-interior solving method is adopted to calibrate the GCM through high-precision GCPs.Finally,the positioning accuracy of less than 2 pixels is achieved,which demonstrates the effectiveness of RGPM and the corresponding calibration method.This research could provide a useful reference for on-orbit geometric processing of long linear whisk-broom remote sensing cameras. |
PDF Full Text Request