Ship Detection using Remote Sensing

This article was originally posted on my firm's website: https://www.mapmyops.com/detecting-ships-in-gulf-of-kutch-and-in-suez-canal and is updated there when an edit needs to be made. Hence, I'd recommend you to read this article from the hyperlink above.

________________________________________________________________

Satellite Imagery can be used for several purposes - climate monitoring and land monitoring being two important areas. In this exercise, I attempted ocean monitoring - ship detection in the Gulf of Kutch using Sentinel 1 satellite imagery with the intention to see the extent of (negative) change in shipping activity in an important maritime region of India due to the lock-down conditions prevalent currently. More importantly, my aim is to demonstrate to the readers, using small and relatable exercises, the scope and benefits of using location intelligence.

The result of the ship detection exercise is depicted in the image below - a more than 1/3rd reduction in shipping activity in the gulf going purely by the count of the objects detected.

(I have used Sentinel 1A GRD data. Much thanks to RUS Copernicus and Open Street Map for the invaluable training material)

How is object detection possible? How precise is it?

The satellite in concern - Sentinel 1A - has a spatial resolution of 5m by 20m. The parameter for objects to be detected was kept at a minimum of 30m. (Keeping it lesser increases the chances of false detections, although even smaller objects can be identified).

The principle behind object detection is how light waves / radar interacts with objects on Earth. In this exercise, a ship appears brighter to the C-SAR instrument on-board Sentinel 1 against the background (water) which is darker. CFAR methodology was used for processing (the same methodology is used to detect icebergs in the colder regions). The extent of precision depends on geological factors, methodology of study, instrument properties and even local knowledge about RoI (Region of Interest).

In essence, it is up to the analyst to glean maximum relevant information from the rich data accumulated by the satellite, using appropriate processing tools.

(Please note, the output of satellite imagery analysis is just plotted on a Google map in the image above. The actual satellite imagery looks like this).

How to interpret these results?

A reduction in maritime activity was expected and ~37% seems plausible to me. Certain aspects to note, however, are-

a) Not all objects are commercial cargo / crude ships. Some can be fishing vessels, port vessels, navy vessels, tourism vessels, pirate vessels, other vessels & objects, or even false detections. However, one can assume that - commercial vessels would be clustered around Adani Port and Kandla Port whereas the other types would be located/clustered elsewhere. With the object size of 30m however, a lot of the objects in the non-commercial category have the potential to be filtered out, though.

All else being constant, the difference in count of objects detected between the two images - 148 objects - also lends more meaning to the output.

b) Similarly, not all objects detected imply movement or commercial activity either. As you would notice, a fair number of ships are clustered around the coastline. These can be docked / dysfunctional / temporarily out of service.

c) This is the comparison of objects detected during only two points in time, one year apart. Adding more time periods i.e. multi-temporal analysis would accentuate the patterns and lend more meaningful output.

What are the possible uses of object detection?

Satellite Imagery when combined with other observation tools are used by Security management professionals to compare the reported whereabouts of vessels (mandatory for ships above a certain size) v/s the actual vessels detected at sea using imagery to spot and track 'dark ships' i.e. those that do not report these location and direction information - these can be be illegal or a threat to national security. Icebergs and Oil Spills can also be detected and monitored using Object detection techniques on water. On land, the purposes are several too. For example, Insurance companies use object detection for risk underwriting. Where else do you think object detection can be useful?

Feel free to share your thoughts as comments or via email at [email protected]. Send in a demo request for commercial grade GIS solutions via email or here. Will be able to respond to requests which can be performed on a remote basis during this lock-down period.

Kind regards, Arpit