X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 13 Feb 2017 02:55 PM by  anon
empirical line calibration
 1 Replies
Sort:
You are not authorized to post a reply.
Author Messages

anon



Basic Member


Posts:228
Basic Member


--
13 Feb 2017 02:55 PM
    Good afternoon, I have a coupleof questions regarding the Empirical Line Calibration tool in ENVI 5.3. I amusing the Empirical Line Compute Factors and Correct tool. Throughout thedialog that being once you run this tool, there are several different optionsthat I do not understand, nor could I find in Help articles online, perhaps Iam just looking in the wrong places but I wanted to ask anyways. Firstly, somebackground on the dataset that I am using. It is an RGBdata set collected from a consumer grade Sony camera from an Unmanned AerialVehicle platform over a forest. Unfortunately, I cannot find out informationregarding Gain or Offset on the sensor, nor do I know the specific wavelengthin which it captures data. While this imagery was being collected, fieldsamples of the reflectance of white and black tarps were taken using aSpectroradiometer. The goal is to use the Empirical Line Calibration techniqueto get reflectance values for the sensor. In theEmpirical Line Compute Factors and Correct tool I have selected the "DataSpectra" to be from Regions of Interest. I digitized the regions ofinterest to match the footprint of the tarps that the field spectra werecollected with the Spectroradiometer. The Data Spectra are in the form of"Digital Numbers" ie. a range from 0-255. For the fielddata I have ASCII files from the Spectroradiometer which have a reflectancereading every nanometre from 350-2000nm. I have attachedsome pictures from the dialog boxes to help show you what I have saidabove. My questionrefers to the number of bands and the wavelengths that are shown in theattached images. For the ImageData, it says there is one band, which I understand to be true as it is onlythe Red band of a true colour image, so that makes sense to me. In the metadatafor that RedBand Image, I went in and manually input a Wavelength of 675nm,which is to represent the middle wavelength of the band I am interested in(600-750nm). However, I cannot seen to put in an additional wavelength value inorder to get a range. Is it possilbe to define the wavelength as a range from600-750nm? In the metadata I added a FWHM of 75nm to try and achieve thisresult but to my knowledge nothing happened. Secondly, is ita problem that my Field Data has 2151 bands and a wavelength from 350-2500?(That is what I think the wavelength for the Image data should also look like,a range from 600-750nm) When I ran thedialog an output was successfully generated, but the result was still indigital numbers when I thought it was supposed to be reflectance. And the graphthat was generated also was not clear to me as it was just a straight linehorizontal to the X axis with only one point on it at 675nm. Any assistanceyou can provide would be greatly appreciated. Thank you for your time,

    Deleted User



    Basic Member


    Posts:228
    Basic Member


    --
    13 Feb 2017 03:03 PM
    If your input image top the empirical line calibration is a one-band image, as you say, then it is expected that in the resulting plot, you will see only one point. This point will be the calibration coefficient to convert values from the image (DN) to the reference (presumably reflectance) spectral space, for that one input band. The output plot is not very useful (in my opinion), especially in the case where you have only one band to consider. There are a couple of confusing things about your message, though. First, I notice in your screen shots that you show four ROIs for the tarps in the image. These are named in a way that it sounds like two of these are dark tarps, and two are white. But for your field spectra, your screen shot shows only one dark spectrum. Are you later adding a white spectrum, and then matching your image and field spectra (dark to dark, light to light)? That is usually how the empirical line calibration works. I am not sure why you want to specify a spectral range between 600 and 750 for the red band of your RGB image. One wavelength, in the center of the red area, would be sufficient in the header of this file. ENVI would not consider a range of values defining the full range of red wavelenths during an empirical line calibration anyway. It would merely pick in the field spectra, the reflectance value for whichever band is closest to the wavelength used for the input image, and base the calibration on that. Best, Peg
    You are not authorized to post a reply.