DOS in Erdas Imagine

One simple Modeller application in Erdas Imagine for the calculation of a DOS-corrected Scene!
Please excuse the bad audio.
you can find out more, visiting this links:
http://www.unc.edu/courses/2008spring/geog/577/001/www/Song01_RSE.pdf
http://www.indiana.edu/~act/files/Dengsheng/atmospheric_correction_IJRS_2002.pdf

Category:

Science & Technology

Tags:

• DOS
• dark object subtraction
• Erdas Imagine
• Modeller
• Modell
• atmospheric correction

[KetamaTV]

Thank you a lot my teacher :)

[ricckli]

@KetamaTV good to hear, that i was able to help!!!

[jbama21]

Thanks for the video, very informative. I am new to erdas and image analysis, your videos help to bridge the learning gap. In your opinion, will I have better classification results if I perform DOS on the scene, as you show, and then perform a supervised or unsupervised classification, in comparison to segmentation or object based analysis. A paper I have been reviewing uses segmentation from decorrelation stretch transformation, 3 tassel cap, and 2 ratio bands. Are you familiar with this?

[ricckli]

@jbama21 thanks for your reply. first of all: the dos-correction does not improve unsuper or supervised classification due to the fact, that you are not changing the overall spectral distance of two pixels. your are just correcting the total image with the way described in the video. decorrelation stretch sound totally different and much more "heavy" regarding the change of the spectral histogramm of an image. I haven´t used object based or segmentation methods. 

[ricckli]

@jbama21 i have read some papers that describe the maximum likelihood classification approach as a simple and very good classifier!!! but it surely depends on your image/quality of the data. what are you dealing with? if you need some help contact me via email! all the best, riccardo

[nadi87s]

What about the hybrid classification methods, I think, By setting both parametric and non parametric factors you will get improved results!

[jbama21]

@jbama21 Thanks a lot for your advice. I meant to add, I've read that in pre-processing the images should go through some contrast stretch. I used a histogram equalization across the layer stack., but the resulting product looked very bad.

[jbama21]

@ricckli I am completing a Masters paper and a small part has been land cover change for the study area. I have two scenes that I need to classify and time is a constraint, so I need to classify the images as accurately as possible but as quick as possible. I performed atmospheric correction for both images, after which i used the mosaic tool to create one image. With time being an issue, would you recommend the max likelihood supervised?