I’m studying the RUS Webinar on ship detection [available here].
The steps-processing performed are:
Subset
Apply Orbit File
Ocean Object Detection
Land-Sea-Mask
Calibration
Adaptive Thresholding
Object-Discrimination
Write
No one Geometric Correction is performed (no TC or EC) in the Webinar. Anyway, is shown that the vector data exported (related to the ship detected) can be visualised in a correct (?) position on QGis.
I would expect huge position errors! What I’m missing?
Just adding to @ABraun’s explanation: ships are moving, so highly accurate position at time of overpass doesn’t add much insight to a ship’s current position. In some areas (Canada’s Bay of Fundy) tides could have an impact. If there is a stationary target like an oil platform in your region you can use that to check your detections.
Thank you very much for the answer Mr @ABraun and Mr @gnwiii .
I’ll perform some tests using GPS ships data positioning to validate the results. I’d like to better understand the position accuracy of S1 GRDH data over the ocean without performing the TC.
As @ABraun said Terrain corrections are required in case of distortions caused by topography such as foreshortening and shadows whereas ocean surfaces can be safely assumed flat.
Moreover, if targets are moving then you have to consider azimuth shifts. Azimuth shifts are caused by relative motion between SAR and target. You can find formula and detailed explanation of azimuth shifts here -
(This is most important step before validation with GPS/AIS data.)
Once, you azimuth corrected your data, then you can use GPS/AIS ship data to match SAR ship detections. Remember, resolution of SAR data has also an effect on your positional accuracy.