X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 20 Dec 2010 07:37 AM by  anon
RapidEye 3A-NAC images
 8 Replies
Sort:
You are not authorized to post a reply.
Author Messages

anon



New Member


Posts:
New Member


--
20 Dec 2010 07:37 AM
    Hi - can anybody offer any advice about processing RapidEye 3A-NAC imagery with ENVI 4.7? I want to generate natural color images and implement a manual linear stretch on each image. Specifically, I'd like to get some advice about the following 2 subjects: 1.) Is it possible to implement atmospheric correction on RapidEye images within ENVI? I understand that I need to reprocess the data with the atmospheric values in the metadata. Typically, I would expect to invoke atmospheric correction through the Basic Tool > Preprocessing > Calibration Tools menu, but none of the options available there seem to apply specifically to RapidEye images. Because the RapidEye images are delivered as a single geotiff containing all 5 data bands, I had already created composite RGB images using ArcGIS before I realized there was an issue with the radiometric values. Is there a way to avoid breaking out the image into its component bands and implement the atmospheric correction in a composite image? I have a lot of images to process and I don't have an IDL license, so it's going to be really painful to separate the bands and then recomposite ... 2.) Can anybody suggest a workflow in ENVI (or another app) for implementing the UDF (Unusable Data Mask) files that are accompany each 3A-NAC RapidEye image? Thanks.

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    21 Dec 2010 11:16 AM
    ENVI doesn't have a specific calibration tool for RapidEye but calibrations are fairly easy to do by hand if you know the equations to use. First, you will need to apply a scale factor (using Band Math) to convert the DNs to radiance as described in the RapidEye documentation: http://www.rapideye.de/up...age_Products_FAQ.pdf The document also describes how to calculate TOA reflectance. If you want to perform atmospheric correction to get reflectance, you can then use something like FLAASH, Quac, or one of the generic tools in ENVI under Calibration utilities such as empirical line (requires field collected data), dark subtraction, log residuals or IARR, or flat field to get relative reflectance.

    Deleted User



    New Member


    Posts:
    New Member


    --
    03 Aug 2011 03:10 AM
    If I can pick up on this thread again... According to RapidEye's FAQ document: "The digital numbers of the RapidEye image pixels represent • absolute calibrated radiance values for non atmospheric corrected images • reflectance values for atmospheric corrected images (currently not offered for delivery)" So if I am reading this correctly, the pixel values already represent radiance values. My related question is the steps required to prepare the imagery for FLAASH. In ENVI 4.8, if you import the L3A imagery via File -> Open External File -> RapidEye and then save to ENVI format, the metadata is imported also. Now, when you import radiance image to FLAASH, you are prompted to select either "Read Array of Scale Factors.." or "Use Single scale factor". Examination of the xml metadata file reveals that the scale factors are the same for all bands: 9.999999776482582e-03 According to RE's product specification: "These scaling factors are applied so that the resultant single DN values correspond to 1/100th of a W/m2 sr-1 μm-1 Now, according to ENVI, FLAASH expects μW/( cm2 • nm • sr). W --> µW: factor of 106 m2 --> cm2: factor of 10-4 µm-1 --> nm: factor of 10-2 sr-1 -> sr: factor of 101 The scale factor is the product of all terms: scale factor = 106 * 10-4 * 10-2 * 101, or 101. = 100 So the scale factor should be 100 ?

    Deleted User



    New Member


    Posts:
    New Member


    --
    03 Aug 2011 03:12 AM
    If I can pick up on this thread again... According to RapidEye's FAQ document: "The digital numbers of the RapidEye image pixels represent • absolute calibrated radiance values for non atmospheric corrected images • reflectance values for atmospheric corrected images (currently not offered for delivery)" So if I am reading this correctly, the pixel values already represent radiance values. My related question is the steps required to prepare the imagery for FLAASH. In ENVI 4.8, if you import the L3A imagery via File -> Open External File -> RapidEye and then save to ENVI format, the metadata is imported also. Now, when you import radiance image to FLAASH, you are prompted to select either "Read Array of Scale Factors.." or "Use Single scale factor". Examination of the xml metadata file reveals that the scale factors are the same for all bands: 9.999999776482582e-03 According to RE's product specification: "These scaling factors are applied so that the resultant single DN values correspond to 1/100th of a W/m2 sr-1 μm-1 Now, according to ENVI, FLAASH expects μW/( cm2 • nm • sr). W --> µW: factor of 106 m2 --> cm2: factor of 10-4 µm-1 --> nm: factor of 10-2 sr-1 -> sr: factor of 101 The scale factor is the product of all terms: scale factor = 106 * 10-4 * 10-2 * 101, or 101. = 100 So the scale factor should be 100 ?

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    03 Aug 2011 09:02 AM
    Most data are delivered as radiance products that have been scaled to integers to save on storage space and transfer time. According to the rapid eye documentation, (http://www.rapideye.de/products/faq.htm) you should multiply the 'radiometricScaleFactor' to the data to get to floating point radiance units of W/m^2*sr*um, which is the most common unit. So I would first apply this scale factor to the data using something like band math or Apply Gain and Offset tool (the scale factor can be considered the 'gain' and there is no offset). The FLAASH scale factor is divided into the data to get um/cm^2*sr*nm, which is a value of 10.

    Deleted User



    New Member


    Posts:
    New Member


    --
    03 Aug 2011 11:42 PM
    Ok, got it. So if the band scaling quoted in the metadata file is 0.00999, to get floating point radiance units, we need to multiply the DN by the inverse of this number 1/0.009999999 = 100 ?

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    04 Aug 2011 08:35 AM
    No, the scaling to get to the published radiance units is separate than the scaling used in FLAASH. I would multiply the data by the scale factor in the metadata, then use a scale factor of 10 when inputting the file to FLAASH.

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    11 Nov 2011 07:58 AM
    Images in DN need to be converted to floating point radiance. You do not convert to TOA reflectance - FLAASH is going to be the atmospheric correct and conversion to reflectance. QUac can take radiance or TOA reflectance. It isn't always necessary to do an atmospheric correction. It is important if you want to compare images from different dates or differerent sensors or make absolute measurements from the data.

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    14 Nov 2011 07:24 AM
    Quac uses a different method but also produces reflectance images. Both FLAASH and Quac scale the data by 10000 and to integers on output to save on disk space. It takes less disk space to store an interger image than a floating point image.
    You are not authorized to post a reply.