A weighted calibration method of interferometric SAR data

Abstract

The accuracy of the digital elevation model (DEM) generated by the interferometric synthetic aperture radar (SAR) partly depends on the accuracy of system parameters, so it is necessary to calibrate the system parameters. The traditional calibration method models the elevation error as a linear function of parameter biases, and solves the biases through the sensitivity equations. This paper presents a weighted calibration method applicable to interferometric SAR data. It introduces the weightings to the sensitivity equations to discriminate the ground control points (GCPs) with different correlation coefficients and locations. This weighted calibration method can improve the DEM accuracy, and this paper illustrates its successful application to airborne interferometric SAR data collected by Institute of Electronics, Chinese Academy of Sciences.

Weighted joint calibration for interferometric SAR

“…Equation ( 11) is equal to (8), but it is more convenient for solving parameter biases. In (11), a layout of scene pairs which have common TPs as shown in Fig. 2 is instanced to describe the error equations about TPs.…”

“…It depends on the method used to solve the least square system, and possibly on some random variables such as initial conditions for certain algorithms [17]. However, we may find a unique solution to (12) by using Moor-Penrose inverse matrix [2,6,11].…”

“…where • denotes the Euclidean norm of vectors. Since noise varies from point to point, the sensitivity equations expressed by (6) of GCPs and by (7) of TPs have different errors [11]. Therefore, it is reasonable to introduce different weights to these equations in the linear least squares problem.…”

“…The weight matrix P is to discriminate GCPs with different coherences and locations. Since noise varies from point to point, the sensitivity equations expressed by ( 6) of GCPs have different errors, which can be written as follows [3,11]:…”

“…In the implementation procedure, it often models the elevation error as a linear function of parameter biases, and get these biases by solving sensitivity equations. Calibration is by nature an iterative technique [6,[11][12][13][14].…”