Radiometric Correction Part I

Exercise #2Name: ______

Geography 475

Digital Image Processing

Radiometric Correction Part I

Due: Feb. 12, 2010

After briefly examining correction of sensor malfunctions, this exercise will guide you through image importand image calibration, which are typical pre-processing steps. These steps are usually required before remotely sensed data are analyzed.

Before beginning the exercise, download the Lab 2 data off the Geography server onto a local hard drive. Change the permissions to allow write access.

Part I: Sensor Malfunctions

One common cause of radiometric error is sensor malfunction. Fortunately, the effects of this type of internal radiometric error are easily reduced. Two types of sensor malfunction are line start error and banding (also called striping). All data and models needed for this exercise are in z:\exercise2 on the GARSL computers.

Start ERDAS Imagine and load the file badlines.img. This AVHRR scene contains line start errors. First, you must determine what lines are bad. Since bad lines appear white, you need to change the color of the inquire cursor. Select Utility/Inquire Color. Then, change the color to something other than white. Click OK. Now, select Utility/Inquire Cursor. Move the inquire cursor over one of the bad lines. Zoom in close to ensure the cursor is properly positioned over the bad line. In the dialog box displaying image coordinates, change the coordinate system from “map” to “file.”

1. Are image rows represented by X or Y coordinates?

Notice that several decimals are shown as part of each coordinate. You only need a whole number to describe the bad line. The line number is given at the center of the row of pixels (use the nearest whole number).

1. What is the line number of the first bad line?

1. Repeat the process for the other bad lines, and record the line number for each.

Close the inquire cursor and fit the image to the viewer. Click Interpreter/Utilities and then Replace Bad Lines. The input file is badlines.img. Save the output file in a folder on a zip disk named for your last name. Create the folder using Windows Explorer if it does not already exist. This will be our working directory. Call the output file fixedlines.img. Now you must enter the bad line numbers. Make sure to enter the lines as a comma separated list (with no spaces). Click OK. After the process finishes, load fixedimage.img into the viewer.

1. Were the bad lines removed? (If not, go back to the original image and check the line coordinates).

1. Briefly explain how the bad line algorithm works (Hint: use lecture notes and view the model by clicking Interpreter/Utilities/Replace Bad Lines and View).

Now load the file tm_striped.img. Change the band combination of this TM scene, which has only the three visible bands included, to view each band separately.

1. Are the effects of banding evident in each band? In which, if any, are they most noticeable?

Click Interpreter/Radiometric Enhancement/Destripe TM Data. The input file is tm_striped.img. Call the output tm_unstriped.img and save it in your working directory. Click OK. When the process completes, load tm_unstriped.img into the viewer.

1. Was the problem completely fixed or simply suppressed?

1. Describe in general terms how the de-striping algorithm works (Hint: use your text and lecture notes).

Part II: Image Import

From this point forward, you will be guided through a typical sequence of pre-processing steps. You will start at the beginning, with image import. Use Windows Explorer to open the directory \tmscene\. You should see 10 files in this directory. Three of the files are in simple text format and can be viewed using Notepad or Wordpad:

064545.SUM = Summary listing of files on the original CD.

LT503002.HD = Image header file (metadata) ANIMPORTANT ONE!

LT503003.HI = Processing steps undertaken at EROS to convert from a raw satellite feed to a readable form.

The other 7 files are binary format, and represent each of Landsat TM’s 7 bands.

Select View/Details.

1. How large (in megabytes) is each image band?

Use Wordpad to open LT503002.HD. This file contains all sorts of useful information. Use it to answer the following:

1. How many bits does each pixel record?

1. How many rows (lines per file) are in each band?

1. How many columns (pixels per line) in each band?

1. What was the scene acquisition date?

1. What was the scene acquisition time (GMT)?

1. What is that in local standard time (LST)? (Hint: The scene is in the U.S. Central time zone.)

Click Import. Change “Type:” to Generic Binary and “Media:” to File. For input, select the Band 1 file (LT503002.I1). Call the output tm_050692.img and save it in your working directory. Click OK. Another dialog box is loaded. The data set you are importing is in band sequential (BSQ) format (one file per image band). Change “Data Format:” to BSQ.

As you noted from the metadata, Landsat TM data is 8-bit. So, leave “Data Type:” set to Unsigned 8 Bit. Next, enter the number of rows and columns (from metadata), and change “# Bands:” to 7. Then, under “BSQ Options” click the check box next to Bands in Multiple Files.

A third dialog box is loaded. Change the directory to \tmscene\. Click on the Band 2 file (LT503002.I2), change “Band #” to 2, and click Set. Then, set the Band 3 file to 3, Band 4 to 4, etc. This tells Imagine what file corresponds to what image band. When all 7 bands are set, click OK. Click OK once more. It will take some time for Imagine to import and stack the data set.

Once the process is complete, load the image into a viewer. Enlarge the viewer and fit the image to it.

Part III: Radiance and Reflectance Calibration

The radiance calibration converts unitless image DNs to a standard value with physical meaning. The calculation requires information about gains and offsets.

Use Wordpad to open LT503002.HD. The header file contains information about gain and offset (bias) for each band. Enter the values into Table 2 below. Note that Band 6 data are not needed (we are working toward calculating reflectance, and Band 6 data are based upon thermal emission).

Table 1: Gains and offsets for 05/09/92 TM scene

Band / Gain / Offset
1
2
3
4
5
7

1. What is the gain-offset equation for calculating radiance?

Keep in mind that the formula simply describes the line that represents a linear relationship between DNs and surface radiance in each band.

Using Imagine’s Modeler, open the model radiance.gmd. Double-click on the table containing Gains. Enter the gain values into the table (set the Band 6 gain to 1). Then, enter bias values into the Bias table (set Band 6 bias to 0).

Click the Red Lightning Bolt. The input is tm_050692.img. Name the output tm_050692_radiance.img (make sure you save it in your working directory).

After processing, open the tm_050692_radiance.img is another viewer.

Use the inquire cursor to examine the pixel values in the radiance image. Note that values are no longer integer DNs. Rather, the radiance image is in a “floating point” format, meaning it contains decimal values (i.e., it is no longer an 8-bit data set).

1. Use Windows Explorer to find the size (in gigabytes) of the radiance image.

Using Imagine’s Modeler, open the model reflectance.gmd. Note that you need to input irradiance values and solar altitude.

1. What was solar altitude (i.e., sun elevation) for this study area at the time of image acquisition? (Check the header file).

Imagine cannot process angles in degrees. Degrees must be converted to radians. The formula for converting degrees to radians is: [(degrees x ) / 180].

1. What is the angle in radians? (show your work).

Enter that value in the solar altitude scalar box in the model.

Next, we must determine exoatmospheric solar irradiance (E). These numbers are published in the article by Chander et al.(2009) posted on Blackboard.The numbers you need are in Table 3 on page 896 of the article. Keep in mind that these data are TM(LPGS) and from Landsat 5. Enter the values in Table 3 below.

Table 2: Exoatmospheric Solar Irradiance for Landsat 5 TM

Band / ESUNλ
1
2
3
4
5
7

Next, enter the values into the table in the radiance model. Make sure to set the value for Band 6 to 1. Click the Red Lightning Bolt. The input is tm_050692_radiance.img. Name the output file tm_050692_reflectance.img and save it in your working directory.

After processing, load the reflectance image into a viewer. Use the inquire cursor to examine pixel values. Note that they should all be decimals. The decimals correspond to the percent of incoming energy that is reflected off each 30 x 30 m patch of ground (i.e., pixel) in each band. Now we’re talking!

But, the file size is huge! We can “fit” the image back into an 8-bit scale by multiplying decimal reflectance values by 10. Load reflectance_8bit.gmd into the modeler. Click the Red Lightning Bolt. Input is tm_050692_reflectance.img. Name the output tm_050692_reflectance_8bit.img and save it in your working directory.

Use the inquire cursor to scroll around the image. Now, pixel values represent percent reflectance in each band.

1. What is the resulting files size (in megabytes) of the 8-bit reflectance image?

Part IIIb: Built-in Tool for Landsat 7 ETM+

Imagine has a built-in tool for calculating reflectance of Landsat 7 ETM+ data. Click on Interpreter/Spectral Enhancement/Landsat 7 Reflectance. The input file is mexico_city_mspec_6band.img. Name the output file mexico_city_reflectance.img. Click on the Conversion tab. Here, you need to enter a variety of information:

A)Make sure the input is set to Raw DN. Solar elevation is in the file L71026047_04720020106_MTL. You can enter the elevation in degrees into this tool. What is solar elevation for this scene? ______

B)To calculate solar distance, find the Acquisition Date in L71026047_04720020106_MTL and then figure out the day of year (DOY). Use DOY to find solar distance on Table 6 of Chander et al.(2009). Enter the value into the conversion tool and write it here: ______

C)The header file L71026047_04720020106_MTL contains LMAX and LMIN values for each band, so change Technique to LMAX/LMIN. Enter LMAX and LMIN values from the header file into the table below and into the Conversion table.

Table 3: LMAX and LMIN for 01/06/02 ETM+ scene

Band / LMAX / LMIN
1
2
3
4
5
7

D)Click OK