Sunteți pe pagina 1din 2

4.

1 Radiometric calibration
Pixel values in commercially available satellite imagery represent the radiance of the
surface in the form of Digital Numbers (DN) which are calibrated to fit a certain range
of values. Sometimes the DN are referred to as the brightness values. Conversion of
DN to absolute radiance values is a necessary procedure for comparative analysis of
several images taken by different sensors (for example, Landsat-2 versus Landsat-5).
Since each sensor has its own calibration parameters used in recording the DN values,
the same DN values in two images taken by two different sensors may represent two
different radiance values.
4.1.1 Main elements of sensor calibration
4.1.1.1 Absolute Radiometric Calibration – from radiance to DN and back
The following figure depicts a hypothetical response curve to a known calibration
signal. There are several common curve characteristics to note. The first is that the
response curve does not pass through the origin, i.e. the sensor does not give zero
output even in the dark. This so-called dark signal is caused primarily by electronic
noise in the sensor. The optimum operating region is the linear region indicated in the
figure. Remote sensing systems are designed to measure the radiometric
characteristics of the targets of primary interest in this linear region, where the output
signal is linearly proportional to the input signal.
Hypothetical calibration response curve, with linear region of slope m and intercept b
This linear function is described by three parameters: the range of DN values in the
image, and the lowest (Lmin) and highest (Lmax) radiances measured by a detector
over the spectral bandwidth of the channel. Most commonly, the data are distributed
in 8-bit format corresponding to 256 DN levels. Lmin is the spectral radiance
57
corresponding to the minimum DN value (usually, a value of 0). Lmax is the radiance
corresponding to the maximum DN (usually, the value of 255).
For each spectral band, the output of the sensor, the grey level or DN, is related to the
input signal, the radiance L in most cases, by the following equation:
DN = G · L + D
The quantity G is the gain of the system and D is the dark current. All of the quantities
in the equation are a function of wavelength but, for simplicity, this dependence is not
indicated in the equations presented here. The image DN recorded by the sensor
system can be converted to the physical quantity L (expressed in Wm-2 sr-1) by
inverting equation:
L = (DN - D) / G = (Lmax-Lmin)/255*DN + Lmin
The information about sensor calibration parameters (Lmin and Lmax) is usually
supplied with the data or is available elsewhere (e.g., Landsat Data User's
Handbook).
If the input signal exceeds the amount for which the sensor was designed, the system
response will become non-linear or reach the saturation level. This is a common
occurrence in land remote sensing systems when they image bright clouds and/or
snow cover, for example.
4.1.1.2 Uniformity Calibration
Many remote sensing sensors in use today produce images. During the calibration
process it is important to determine how the system responds to signals coming from
different parts of the scene. The optics of a sensor system will be a determining factor.
A phenomenon know as vignetting will reduce the light reaching the detector system
towards the edges of the lens. Similarly, many sensor systems use arrays of detectors
to image scenes more quickly. The detectors involved always have some differences
in response to the same input signal level, which can lead to image striping. Thus, to
achieve uniformity in response across the field of view of the sensor, relative
radiometric calibration is also necessary.
4.1.1.3 Spectral Calibration
To determine a sensor's response to light at different wavelengths, an input signal of a
known wavelength and intensity is scanned across the wavelength range and the
response as a function of wavelength measured and characterized for each spectral
band prior to launch (see the figure). If operational applications are not to be
compromised, it is important to know the position and width of each spectral band, as
well as to understand out-of-band contributions from other spectral bands.
58
Relative spectral response profile for a broad-band sensor
4.1.1.4 Geometric Calibration
Due to optical aberrations or misalignment of discrete multiple detectors, images from
different spectral bands or detectors can be misregistered. In a framing sensor, straight
lines may appear curved. In this context the known input would be a test pattern of
lines and geometrical shapes. The resulting output image would be compared to the
input image and the appropriate geometric calibration equations implemented

S-ar putea să vă placă și