DSTA, Singapore navy develop AI-powered vessel-classifying tool

A screen grab of the AI-driven computer vision system, which has been developed by DSTA in conjunction with the RSN. (DSTA)

Singapore's Defence Science and Technology Agency (DSTA) has partnered with the Republic of Singapore Navy (RSN) to develop a computer vision-based command, control, and communications (C3) system that uses artificial intelligence (AI) to detect, monitor, and classify vessels in real time.

This initiative, which gather information via multiple sources including optical cameras and automatic identification system (AIS) data, is part of a broader effort to enhance Singapore's maritime situational awareness amid increasingly complex security challenges.

In a 16 September demonstration of the system, which was attended by Janes, officials from DSTA described the computer vision system as one that can classify a vessel in less than one second. 

In contrast, methods that rely heavily on human operators visually identifying vessels can take up to 20 seconds per vessel, a DSTA representative said.

To achieve the speeds at which it can classify the vessels, DSTA's computer vision system combines video analytics with deep machine learning.

Models running video analytics were fed with images of various vessel types that range from large tankers to small boats and are trained to infer a vessel's class based on features such as its overall length, size, and other visual attributes.

The system is also designed to alert operators only when anomalies are detected, such as when a vessel is berthed suspiciously along a large tanker, which could denote illegal ship-to-ship transfer activities.

This AI-driven sifting of potential anomalies allows operators to focus on high-priority threats.

To ensure accuracy, DSTA engineers worked closely with RSN domain experts to collect and label thousands of operational images, the DSTA representative said.