Logo GISdevelopment.net

GISdevelopment > Proceedings > ACRS > 2004


1989 | 1990 | 1991 | 1992 | 1994 | 1995 | 1996 | 1997 | 1998 | 1999 | 2000 | 2002 | 2004
Sessions

New Generation Sensors and Applications

Hyperspectral Sensing

Application of New Sensors

Airborne Sensing

3 Line Scanner

LiDAR

Digital Camera

New Generation Sensors

Data Processing

DEM/3D Generation

Change Detection

Data Fusion

Hyperspectral Data Processing

Automatic Feature Extraction

Automatic Classification

High Resolution Data Processing

Data Fusion

Image Classification

High Resolution Data Processing

GPS & Photogrammetry

Navigation System

Digital Photogrammetry



ACRS 2004


Data Processing: Hyperspectral Data Processing


On The Atmospheric Correction for a Hyperion Scene



THE ATMOSPHERIC CORRECTION
There are three files coming with the distributed scene, the metadata (.met), the image (.L1R), and the header file (.hdr). The image file is stored in BIL order. For the atmospheric correction, the scene center location is required. This information is available from the .met file. The .met file is a short file. For the scene obtained, the .met file contents are listed as in the appendix A. Besides the scene center locations, the four image corners are also provided with coordinates.

The image used in this study is EO1H1170452003240110KZ. “EO1” stands for the satellite EO1 and “H” stands for Hyperion. The numbers, 117 and 045, are the WRS path and row respectively. 2003 is the year of image acquisition while 240 is the Julian day of acquisition. The first “1” following “240” indicates that the Hyperion sensor is on. The second “1” indicates that the ALI sensor is on. The following “0”, indicates that the AC sensor is off. “K” is a code for the pointing mode, and “Z” is the code for the scene length.

For FLAASH, the flight data and time in GMT is also required. In the header file, there is a date statement, description = { Hyperion L1 Data Product [Thu Sep 11 01:04:52 2003]} The date listed on the .met file and image title for day 240 is August 28, rather than September 11. The one in the .met file listed as HYP start time is used. The other parameter required is the flying height and ground height in kilometers. The flying height is 705 KM from the EO-1 satellite description. The ground height is set to zero km For the scene area in TaiDong (Site Latitude, 22.060000; Site Longitude, 120.730000).

There are 220 unique spectral channels collected with a complete spectrum covering from 357 - 2576 nm. The Level 1 Radiometric product has a total of 242 bands but only 198 bands are calibrated. Calibrated channels are 8-57 for the VNIR, and 77-224 for the SWIR. The reason for not calibrating all 242 channels is mainly due to the detectors' low responsivity. The bands that are not calibrated are set to zero in those channels. There are only 196 unique channels (USGS, 2004b), because of an overlap between the VNIR, band 56 (915.23 nm), 57 (925.41 nm) and SWIR, band 77 (912.45 nm), 78 (922.54 nm), focal planes. In the experiment, all those un-calibrated bands and band 77 and 78 are removed before further processing. The digital values of the Level 1product are 16-bit radiances and are stored as a 16-bit signed integer. The SWIR bands have a scaling factor of 80 and the VNIR bands have a scaling factor of 40 applied. The units are W/m2 SRµm (USGS, 2004b).

VNIR L = Digital Number / 40
SWIR L = Digital Number / 80

An ASCII file is prepared for each band, 196 lines corresponding to 196 bands. The first 50 lines are 40 and the next 146 lines are 80. This file is saved as gain_invmicro.txt. From ENVI, the original Hyperion file could be read directly, but for the atmospheric correction, the native file should be used. The saved file of the ENVI native file is in BSQ format. This has to be converted to either BIL or BIP format for the FLAASH program. This task is performed with the function in Basic Tools/Convert Data. There is no need to perform above stated procedures, because the un-calibrated and overlapped bands are removed first. The spectral subset function can be performed with the original .L1R file, and stored in the native ENVI BIL format.

In the FLAASH menu, the scale factor file is specified at the same time when specifying the input radiance file. That is, the gain_invmicro.txt file is specified as the scale factor. The atmospheric model selected is the Tropical, and the aerosol model is Rural. Aerosol retrieval function is turned on and the initial value for the visibility is set to 40 km. The spectral polishing option has been tested for both on and off options. When the spectral polishing is on, the width for the polishing operator is set to be 9. The IFOV value for Hyperion is 0.043. The aerosol scale height and CO2 mixing ratio is using the default value, which is 2.0km and 390 ppm respectively. MODTRAN resolution can be 15, 5, or 1.

The visibility computed is 68.4373 km for taking MODTRAN resolution 15, and the average water content is 4.0794 cm. For MODTRAN resolution 5, the visibility remains the same and the average water content is 3.8722 cm. As shown in Table 1, while the visibility unaffected by the MODTRAN resolution, the average water content is reducing with the increase of the resolution. This is somehow unexpected.

Table 1: The visibility and average water content

After the atmospheric correction, the reflectance value of several bands is too low. The resulting value is set to zero by the software. These bands are band 92- 102 and 104, and 137- 153 in the 196-band sequence number. In the original Hyperion sequence, these bands are, band 101-111, 113, and 146-162. All these bands are SWIR bands. In total, there are 29 bands that contain no value. Therefore, 167 bands are left for further processing.

A portion of the original image displayed in band combination (30, 20, 10) for RGB is shown in Figure 3(left). The center scene of Figure 1 is the MODTRAN resolution 15, spectral polished. The unpolished is shown in Figure 1, right. From Figure 1 to Figure 8, it can be observed that the atmospheric correction has restored the reflectance value well. The reflectance features, such as the linear slope in the VIR band for soil, and the steep rise from red to IR of vegetation, are very illustrative from these spectra.


Figure 1: The upper portion of the scene, radiance (left), spectral polished reflectance (center), spectral unpolished reflectance (right)


Figure 2: Vegetation, the spectral of the center pixel in the red box of Figure 1, radiance (left), spectral polished reflectance (center), spectral unpolished reflectance (right)


Figure 3: The human settlement portion of the scene, radiance (left), spectral polished reflectance (center), spectral unpolished reflectance (right)


Figure 4: Cloud, the spectral of the center pixel in the red box of Figure 3, radiance (left), spectral polished reflectance (center), spectral unpolished reflectance (right)


Figure 5: The inland pond (left), the sea water (center and right)


Figure 6: Inland Pond (top), Sea Water, the spectral of the center pixel in the red box of Figure 5 (right), radiance (left), spectral polished reflectance (center), spectral unpolished reflectance (right)


Figure 7: The bare soil


Figure 8: The bare soil, the spectral of the center pixel in the red box of Figure 7 (top for the left), radiance (left), spectral polished reflectance (center), spectral unpolished reflectance (right)

Page 2 of 4
| Previous | Next |

Applications | Technology | Policy | History | News | Tenders | Events | Interviews | Career | Companies | Country Pages | Books | Publications | Education | Glossary | Tutorials | Downloads | Site Map | Subscribe | GIS@development Magazine | Updates | Guest Book