[ic350450]

Hi All. I am working with a set 63 Sentinel-1A images (plus 1 Sentinel-1B image), ranging from 2015 to 2017, HH polarization, to detect seasonal deformation trends of bridges in Illinois. I have used the “smart = 5” option to monitor nonlinear deformations expected from the thermal effects on the bridges. Attached is a picture of the parameters used to process the images in APS module. The results of the APS show the attached coherence plot for each image. I have attempted several cases including linear models and thermal expansion estimation without the smart option. All of my attempts resulted in APS coherence from “test” option similar to the one attached, with the minor difference in that the coherence was greater for my first linear model since the parameters I used were more stringent. The low coherence images are still low in all cases.

1. Are these images effecting my results?
2. How can I improve the APS estimation? Would eliminating the low coherence images from the dataset help in the APS estimation?
3. When using the Smart option for the purposes of my study (Bridge monitoring using Sentinel-1 Images), should I consider downloading more images or is 64 images sufficient?

I appreciate any suggestions.
-Isaias

Attachments:

You must be logged in to view attached files.

- Isaias

[antig]

Ho big is Your bridge. If it is less than km You may skip APS and use small area module

[periz]

From your attachments, the best result is the linear one.

Few hints that work in general (but might not work in some particular cases):
1. when estimating the APS, better to focus on good scatterers, not affected by thermal or non-linear movement. so, using a linear model only is a good choice. you will include other more complex scatteters later in the MISP module
2. do not use at the same time the smart mode and linear and/or thermal estimation, they might affect each other
3. if you expect thermal expansion, it is better to estimate the correlation with temperature instead of using the smart mode. The smart function applies a smoothing in time, that can follow seasonal trends, but temperatures do also have a bit of high frequency fluctuations.

Coming to your questions:
1. images with low coherence look like being well detected by the algorithm. they should not affect the estimation of your parameters, they will just lower a bit the temporal coherence. you can investigate why those images have low coherence: snowy/thunderstormy days?
2. as said above, the linear analysis look good
3. 64 images is a good number. anyway, the more images you have, the more stable will be the result, in particular if you need to consider yearly signals

The suggestion from Anti clearly can be applied on bridges one by one. you can use it as a cross-check when you have doubts about the reliability of the APS analysis.

best

[Author]

Posts