OKI Develops AI-Based Ship Classification System
A new AI system uses deep learning to classify ships based on their acoustic frequencies automatically. Japanese information and communications company OKI has developed an AI ship classification system that can automatically classify ships through deep learning of underwater sounds.
The technology is able to “continuously and automatically acquire” ship classification data, even in areas where visual identification is difficult, such as ports with heavy ship traffic or at night.
OKI said internal validation tests conducted by the company showed that the system can classify ships with 90% or higher accuracy using only minimal training data.
Sound waves can travel thousands of kilometers underwater, becoming the primary means of classifying seafloor objects due to the weakness of radio waves and the scattering of light waves.
Underwater sounds are “unique” to each source, whether it is an animal or a ship, providing a method of identification. Traditional methods involve manually analyzing these sounds, but the results can vary depending on each person’s skill level.
OKI has developed a system that analyzes the characteristics of sounds picked up by underwater microphones, drawing on its expertise in underwater acoustic research.
How well do you really know your competitors?
The new AI ship classification technology uses deep learning to automatically classify ships based on their sound frequency characteristics, reducing reliance on human expertise and meeting the need for a labor-saving solution.
OKI’s verification process uses ship audio data to create a deep learning model.
Deep learning models typically require “big data” to accurately identify sound types. However, given the limited nature of underwater acoustic data and the challenges of preparing acoustic information on a variety of ships, OKI implemented data augmentation and semi-supervised learning techniques.
The company said these techniques enable the system to achieve “high accuracy” even with limited training data.
“In the future, we will seek partners to jointly conduct field data collection and conduct actual verification in the hope of commercializing this technology,” said Yuichi Kato, head of OKI Toki’s system department and senior executive officer.
