search

isce-framework / isce2

InSAR Scientific Computing Environment version 2
Other
505 stars 249 forks source link

alosStack estimated interferogram coherence larger than 1 #847

Closed ehavazli closed 4 months ago

ehavazli commented 4 months ago

I am processing ALOS-2 ScanSAR data and I ran into a case where the estimated interferogram coherence exceeds 1:

Band 1 Block=5129x1 Type=Float32, ColorInterp=Undefined
  Minimum=0.000, Maximum=897447.438, Mean=8561.947, StdDev=5409.711
  Metadata:
    STATISTICS_MAXIMUM=897447.4375
    STATISTICS_MEAN=8561.9472832599
    STATISTICS_MINIMUM=0
    STATISTICS_STDDEV=5409.7114304967
    STATISTICS_VALID_PERCENT=100
Band 2 Block=5129x1 Type=Float32, ColorInterp=Undefined
  Minimum=0.000, Maximum=1.201, Mean=0.487, StdDev=0.237
  Metadata:
    STATISTICS_MAXIMUM=1.2005478143692
    STATISTICS_MEAN=0.48689489981335
    STATISTICS_MINIMUM=0
    STATISTICS_STDDEV=0.23687659024294
    STATISTICS_VALID_PERCENT=100

The .cor.xml file definition shows: imageMath.py -e=sqrt(b_0*b_1);abs(a)/(b_0+(b_0==0))/(b_1+(b_1==0))*(b_0!=0)*(b_1!=0) --a=diff_221202-221216_5rlks_28alks.int --b=221202-221216_5rlks_28alks.amp -o 221202-221216_5rlks_28alks.cor -t float -s BIL

I think the coherence estimation formula is missing a normalization step but any ideas @CunrenLiang ?

CunrenLiang commented 4 months ago

Yes, that’s because of floating point number computation, which makes the result sometimes slightly larger than 1.

You can make those values smaller than 1, e.g. 0.999, without making another copy of the coherence file by using codes like:

https://github.com/isce-framework/isce2/blob/main/components/isceobj/Alos2Proc/runUnwrapSnaphu.py#L87-L90

From: @.> on behalf of Emre Havazli @.> Reply-To: isce-framework/isce2 @.> Date: Thursday, May 23, 2024 at 9:43 AM To: isce-framework/isce2 @.> Cc: Cunren Liang @.>, Mention @.> Subject: [isce-framework/isce2] alosStack estimated interferogram coherence larger than 1 (Issue #847)

I am processing ALOS-2 ScanSAR data and I ran into a case where the estimated interferogram coherence exceeds 1: Band 1 Block=5129x1 Type=Float32, ColorInterp=Undefined   Minimum=0.000, Maximum=897447.438, Mean=8561.947, StdDev=5409.711   Metadata:     STATISTICS_MAXIMUM=897447.4375     STATISTICS_MEAN=8561.9472832599     STATISTICS_MINIMUM=0     STATISTICS_STDDEV=5409.7114304967     STATISTICS_VALID_PERCENT=100 Band 2 Block=5129x1 Type=Float32, ColorInterp=Undefined   Minimum=0.000, Maximum=1.201, Mean=0.487, StdDev=0.237   Metadata:     STATISTICS_MAXIMUM=1.2005478143692     STATISTICS_MEAN=0.48689489981335     STATISTICS_MINIMUM=0     STATISTICS_STDDEV=0.23687659024294     STATISTICS_VALID_PERCENT=100 The .cor.xml file definition shows: imageMath.py -e=sqrt(b_0b_1);abs(a)/(b_0+(b_0==0))/(b_1+(b_1==0))(b_0!=0)*(b_1!=0) --a=diff_221202-221216_5rlks_28alks.int --b=221202-221216_5rlks_28alks.amp -o 221202-221216_5rlks_28alks.cor -t float -s BIL

I think the coherence estimation formula is missing a normalization step but any ideas @CunrenLiang ?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>