20 May 2015 05:05 AM |
|
while performing target detection using thor target detection for single file tool I am facing an error "SVDC:SVDcmp_d did not converge" and before opening viewer its giving "no signature matching routine has completed successfully" and its not generating any rule image.
please help me to resolve this problem
|
|
|
|
MariM Veteran Member
Posts:2396  
20 May 2015 11:58 AM |
|
The error means there was a failure in the calculation of the signature matching, typically when using an algorithm such as SID or matched filter. SVDC requires a matrix inversion and can fail for many reasons. I would check the statistics of your input dataset for bands that contain all 0s or NaNs. You can specify which bands are bad in the image header file so that these bands are ignored during processing. Or you can do a spectral subset and exclude them.
|
|
|
|
Deleted User New Member
Posts:  
20 May 2015 11:53 PM |
|
Thanks MariM..
ya even from help for some errors I got same reasoning that it might be due to NaNs..
but further how to identify bad bands from image. I tried to run statistics n sum bad bands in envi 5. but unable to interpret its result.
can you please suggest ways to identify bad bands.
|
|
|
|
Deleted User New Member
Posts:  
21 May 2015 12:58 AM |
|
I executed that bad data mitigation option and tool is generating results. I am yet to explore the correctness of these results. but that problem is solved..
is there any way to know bad/faulty bands so these can be permanently removed from HSI image. ( I am new in the field so inquisitive to understand it more)
thanks for the prompt response MariM.
|
|
|
|
MariM Veteran Member
Posts:2396  
21 May 2015 06:35 AM |
|
You can use a tool such as Resize data to create a spectral subset to remove bad bands. The THOR Zero Bands tool can help you find bad bands.
|
|
|
|