| CPC G06V 20/13 (2022.01) [G06V 20/194 (2022.01)] | 11 Claims |
|
1. A method of on-orbit geometric calibration (GC) for a spaceborne segmented linear-array camera based on a rational polynomial coefficient (RPC) model, the method comprising:
using ground control point (GCP) identified from a reference image of a calibration site as observations, constructing a GC model based on current GC parameters and the RPC, detecting and eliminating gross GCPs using a global model error correction, and achieving an accurate solution through a stepwise absolute and relative GC assisted by angular resolutions for the segmented linear-array camera,
wherein the method further comprising:
1) Selecting a satellite remote sensing image for GC, according to an area where the reference image is located, and matching corresponding image points as GCPs;
2) constructing a camera GC model suitable for a geometric processing of optical satellite images using a rigorous physical model (RPM) based on a viewing-angle of charge-coupled device (CCD) detectors, and introducing multiple sets of polynomials to fit viewing-angles of segmented linear-array CCDs, and then determining GC parameters to be solved;
3) Before calculating the GC parameters, according to a correlation between interior orientation parameters (IOPs) and exterior orientation parameters (EOPs), correcting EOP and IOP errors in the RPC by introducing an integrated error correction model of the same order as a viewing-angle model into the RPC, thus reflecting a matching error of the GCPs, which are coupled with the EOP and IOP errors, calculating modified GCP residuals using the error correction model and a threshold defined as three times a root mean square error (3×RMSE) of the residuals, and detecting and eliminating the gross GCPs through an iterative estimation of the error correction model;
4) Based on geometric characteristics of the camera EOPs, constructing an EOP error correction model for RPC with additional translation and rotation transformations, estimating and correcting the EOP errors in the RPC based on the GCPs, and then determining virtual image points that reflects geometric errors of the IOPs;
5) Taking the real and virtual image points of GCPs as observations, taking a condition that the viewing-angles of the virtual and the real image point are equal as constraints, and then on this basis, constructing an adjustment model of the camera GC by introducing angular resolutions;
6) Based on the constructed adjustment model, with assistances of angular resolutions along and across the CCD, performing the absolute GC of each CCD one by one using a stepwise optimization in two directions along and across the CCD, and then compensating absolute geometric distortion of each CCD;
7) Selecting a CCD as the reference, performing the relative GC of non-reference CCDs relative to the reference based on the angular resolutions and EOP corrections, and compensating relative position residuals of each non-reference CCDs relative to the reference CCD; and
8) According to the corrected relative position residuals, based on an imaging inclination angle of the satellite-borne camera along satellite's orbit, constructing a compensation model for bias field-of-view (FOV) distortions, and fusing the compensation of this distortion of each CCD into the calculated viewing-angle model through an overall least squares adjustment, so as to correct the bias FOV distortion along the CCD caused by shifting the non-reference CCD along the orbit.
|