Empty Ocean Object Detections Layer - SNAP 6


#1

I am using the Ocean Object Detection application and I get different behaviour, one method is similar to all the blogs and descriptions and seems to result in an empty layer.

  1. Using SLC, I apply the orbit, subset, calibrate and deburst. Then I apply the ocean object detection to Sigma0. The output has a detections layer under “Vector”

  2. Following the examples that I have seen online and to try to avoid issues with multiple detections of the same vessel I have tried to convert the data to GRD (using the SNAP built-in), I have then tried running the Ocean object detection on this Sigma0 but the output format is different. There is no detections layer in the output and if I try to add the detections layer with the manager as shown in tutorials, it is empty.

I’d be grateful for some explanation of why one works and not the other.

Thank you.


#2

why don’t you just directly download the GRD product? It is totally sufficient for the Ocean Object Detection.


#3

Because I wanted to be able to:
a) test at full res
b) have freedom to select the most appropriate multilooking for my needs
c) I need the SLC for other applications so there’s no point in downloading both SLC and GRD when I can easily create one.

However, as discussed in my question I did create a GRD in SNAP and tried running the detection on it but it seems to output an empty layer when I follow these instructions:

However when I create the Sigma0 from the SLC the ship detections are listed under “vector”.

I hope I’ve sufficiently answered your question regarding why I am not downloading the GRD.


#4

I see, thank you for clarification.
The only reasons I could think of are that the object size which is different for GRD and the masking process of land area.


#5

I don’t think size can explain the empty layer or different product content that you can see in the attached images.

The long named product is the product with the detections derived from the SLC. You can quite clearly see there is a vector layer of detections and the product content is different to that for the product described below.

The other is the SNAP generated GRD which shows no vector data for the detections and there is no masks container.

If we look at a histogram of the data, it is quite clear one is empty, the other not.

First, the detections bit mask from the SLC data and second the detections from the GRD data.