GEOG 751: Digital Techniques of Remote Sensing

Exercise #6

Radiometric Correction Using Empirical Line Calibration

Objectives

Part I - Empirical Line Calibration

The following description of Empirical Line Calibration (ELC) is directly from the text by Dr. Jensen (2004).  If you want more specific knowledge of ELC, please refer to Jensen (2004): pp 210-215. 

Atmospheric correction can be performed using Empirical Line Calibration (ELC).  ELC has the remote sensing data match in situ spectral reflectance measurements, which are obtained at approximately the same time as the remote sensing overflight.  ELC is based on the equation:

 REFLECTANCEk = Ak * BVk + Bk,

where BVk is the digital brightness value for a pixel of band k, REFLECTANCEk equals the in situ surface reflectance of the materials within the remote sensor IFOV at a specific wavelength, Ak is a multiplicative term (gain) affecting the BVk, and Bk is an additive term (offset).  The multiplicative term is associated primarily with atmospheric transmittance and instrumental factors, and the additive term deals primarily with atmospheric path radiance and instrumental offset (i.e., dark current).

To use ELC, the analyst usually selects two or more areas in the scene with different albedos (e.g., one bright target such as a sand pile and one dark target such as a deep, nonturbid water body).  The areas should be as homogeneous as possible.  In situ spectroradiometer measurements of these targets are made on the ground.  The in situ and remote sensing–derived spectra are regressed and gain and offset values computed. The gain and offset values are then applied to the remote sensor data on a band by band basis, removing atmospheric attenuation. Note that the correction is applied band by band and not pixel by pixel.

Most multispectral remote sensing datasets can be calibrated using empirical line calibration.  The difficulty arises when trying to locate homogeneous bright and dark targets in the study, collecting representative in situ spectroradiometer measurements, and extracting uncontaminated pixels of the calibration targets from the imagery.  If the analyst does not have access to in situ spectra obtained at the time of the remote sensing overflight, it might be possible to use spectra of such fundamental materials as clear water and sand (quartz) that are stored in spectral libraries (e.g., from NASA’s Jet Propulsion Laboratory (JPL), USGS, Johns Hopkins University spectral library).  Hopefully, some of these materials exist in the scene and the analyst can locate the appropriate pixel and pair the image brightness values with the library in situ spectroradiometer data.

You will use several spectral pairs to perform the Empirical Line Calibration.  Those pairs are used to plot a regression line, which is used to modify the input image.  Since the purpose of the ELC is to define the regression line, you should use at least one pair of spectra from both bright and dark areas. 

 

Part II - Perform Empirical Line Calibration Using the IMAGINE Spectral Analysis

Image


Spectral Library

In this exercise, we don’t have in situ reflectance measurements, which were obtained at approximately the same time as the remote sensing overflight.  Thus, we will use some spectral measurements from the spectral libraries.  Open the charleston.img in the new viewer, and check there are sea water and sand beach.  We will use samples of sea water as dark targets and samples of sand (quartz) beach as bright targets.  ERDAS IMAGINE doesn’t have a spectral library of sea water for visible and NIR spectral regions.  Before start, please make copies of two library files (coastseawater.txt and quartz.txt) to save in the folder (C:\\Program Files\IMAGINE 8.7\etc\spectra\aster\user\).  You need to create the folder of “user.”

Start Spectral Analysis (Empirical Line Calibration)

Now you will start empirical line calibration.  Click the Classifier icon on the ERDAS IMAGINE icon panel.  Select Spectral Analysis from the Classification menu.  Then, you can see the Spectral Analysis Workstation button on the Spectral Analysis menu.  When you click the button, you will get a dialog, which looks like this:

 

Now you will add the image to perform the ELC.  In the Spectral Analysis Workstation, select [File | Open Analysis Image].  Navigate to the folder where you saved the dataset of EX06, and select the file Charleston.img, then click OK in the File Selector.  Then, you will get a Sensor Information Query window.  Select a sensor as Landsat TM – 6 bands and click OK.  The image that you are looking at has not an appropriate color composite for the exercise.  Let’s change it with a false color composite (RGB=432).  Right-click in the Main View and select Arrange Layers from the Quick View menu.  In the Arrange Layers dialog, right-click on the layer Charleston.img.  Select Band Combinations from the TrueColor Options menu.  Set the Red, Green, and Blue display to bands 4, 3, and 2, respectively.  Click Apply, then Close in the Set Layer Combinations dialog.

 

Open Library Files

Now you will identify the two library files that you saved before.  In the Spectral Analysis Workstation, click the Atmospheric Adjustment icon ().  You get another dialog window, which looks like this:

Click the Method popup list and select Empirical Line.  You can see the computation window showing slopes, intercepts, and errors per band.  In the Spectrum Library window of the Atmospheric Adjustment Tool dialog, expand ASTER Library (click + symbol next to ASTER).  It takes some time.  When the expansion is completed, you can identify two spectra (coast sea water and Quartz Milky TS-1D).

 

Identification of Samples and Reference

First, you will collect the samples of dark target (Sea Water).  Adjust the position of the Link Box in the Main View so that Ocean is visible in the Zoom View.  Click the Color Chooser icon () and select the color Blue.  Click the Point AOI icon () in the AOI tools.  Click inside the area of Ocean water, which displays as fairly a uniform blue feature near the bottom right corner of the image (e.g., coordinate (682, 836)).  You can see the spectrum of the pixel displays in the Spectral Plot at the bottom of the Workstation, and is labeled Sample.  Now you will add the reference spectrum.  Click the Spectrum symbol to select Coast Sea Water and drag the Coast Sea Water spectrum to the Spectral Plot.  The spectrum appears labeled as Reference. Please collect another Sea Water sample (e.g., coordinates (707, 842)) using the same procedure.

Now you will add the samples of bright target (Sand Beach).  Adjust the position of the Link Box in the Main View so that the Sand Beach in the middle right of the image is visible in the Zoom View.  Click the Color Chooser icon and select the color Red.  Click the Point AOI icon in the AOI tools.  Click inside the area of the Sand Beach, which displays as fairly a uniform white feature near the middle right part of the image (e.g., coordinate (729, 489)).  You can see the spectrum of the pixel displays in the Spectral Plot at the bottom of the Workstation, and is labeled Sample.  Click the Spectrum symbol to select Quartz Milky TS 1D.  Drag the Quartz Milky TS 1D spectrum to the Spectral Plot.  The spectrum appears labeled as Reference.  Collect another Sand Beach sample (e.g., coordinate (689, 526)) using the same procedure.

 

Perform the Radiometric Correction (ELC)

You can see the calculated regression (slope/intercept) line for each band.  It is also possible to review the spectral pairs by clicking the Up Arrow icon and the Down Arrow icon.  If one point seems far off the line, perhaps you can remove it using the Delete icon and collect samples again.  Please save the calculated regression information as “ELC_coefficients_YourName.aad” and leave a copy of it in your Assistant folder.  In the Spectral Analysis Workstation, click on [View | Preprocess | Atmospheric Adjustment]. This processes the analysis image through the selected preprocessing step (Atmospheric Adjustment), and displays the processed image in the Spectral Analysis Workstation.  If you can’t find Atmospheric Adjustment in your preprocessing menu, please make it sure that the Use is checked in Atmospheric Adjustment Tool.  Note that the resulting image is a temporary file – if you wish to retain this intermediate image, you must save it.  Select [File | Save Preprocessed Image].  Navigate to your folder and name the file charleston_elc.img, then click OK in the File Selector.

 

Compare the Calibrated Image with the Original One

Now you will compare the Empirical Line Calibrated image with the original TM data.  Please close all Spectral Analysis windows.  Open two images (original TM image: charleston.img and Empirical Line Calibrated image: charleston_elc.img) in the two new viewers.  Click profile icon and select Spectral Profile for each viewer.  Using Point Profile icon, click somewhere in the viewer and set the coordinate as (312, 29) for both viewers (one for original TM image and the other for empirical line calibrated image).  Healthy vegetation covers that location.  Using another Point Profile icon, place it in the pixel of Sea Water that you think it is representative.  With the same procedure, create Point Profiles of Urban area for both images.

You can see the profiles more clearly when you modify the profile chart option for Y axis (like this - the title: reflectance, min: -0.1, max: 1.0, Major Incr.: 0.1, and Format: 0.0).

 

Questions:

1.  Compare the spectral profiles between the ELC image and the original image.

- Healthy Vegetation

- Urban

- Sea Water

2.  Do you think the ELC method corrected the atmospheric effect very well?  What errors might be included in your ELC method?  Evaluate your ELC result.

3.  How much do you have negative reflectance in your ELC image?  How about the reflectance larger than 1?  Which features have the negative reflectance or the reflectance larger than 1 per band (you might want to use spatial modeler)?  Why do you think your result has the reflectance less than 0 or larger than 1?  How can you fix the problem?


Part III - Create a Map with Your Results

Now use your creative abilities to create a map (exercise-6.map) to compare the two images (the original one and the ELC image).  You may want to display and compare the spectral profiles... the choice is yours. Be sure to include map essential elements that are appropriate for your image.  Save the map composition in your Assistant directory when finished.

References

John R. Jensen, 2004, Introductory Digital Image Processing, 3rd Ed., Upper Saddle River, NJ: Prentice Hall, 526 pages.

Leica Geosystems, 2003, IMAGINE Spectral Analysis User's Guide, Atlanta, GA.



Back to GEOG 751 Home Page

Last Modified: January 16, 2006