java.lang.NullPointerException - Object Ocean Detection

Dear all,

I’m facing a strange problem when trying to detect ships on Sentinel-1 data.
If I import my data, calibrate it and extract ships using RADAR > SAR Application > Ocean Application > Ocean Object Detection, eveything runs smoothly, but if I add an ellipsoid correction step before the object detection (using Radar > Geometric > Ellipsoid correction > Average Height Range Doppler for example) the object detection crashes (java.lang.NullPointerExecption).
I first suspected a problem related to “no data” values, but the ojject detection keeps crashing after removing “no data” values et remplacing them by “0”.
There is nothing related to the crash in .snap/system/var/log/messages.log

Did anyone already face such a problem / have some hints that could help me to solve my problem?

Thanks :slight_smile:

why do you want to apply the ellipsoid correction to your image before the object detection?

You can as well do it afterwards or get the coordinates of the objects from the log file?

Thanks!
The fact is that I observe descrepencies between the object detection output file (xml in /var/log) and the image.
For example, if I visualize the detected objects, pick one, get its coordinates and look back in the xml file in /var/log I cannot find the extracted coordinates there. Therefore I suspected that the results in the xml output could be expressed in a different coordinate system compared to the original image. (My goal is then to crop the image at the detected objets positions using a custom script taking as inputs the xml log and the image)
Now, If I extract the object and then make the ellipsoid correction the problem is the same.
I which coordinate system are the object detection output expressed?

alright, I see…

The cordinates from the logfile are in WGS84 (EPSG:4326) and can be converted into a csv file with a few steps.

Depends on your data how you would search for the possible error.
a) Geocode the resulting raster and see if it matches?
b) Export the logfile and check in a GIS. I received quite good results with it:

This is how the logfile looks:

And this is how the csv fils should look like

Afterwards the csv can be imported into GIS

:

Thanks! I will work on that and check if I can achieve my goal.