Veluon

Sonar Object Detection

Software engineeringAI & machine learningScientific R&DData science

Veluon developed an object-detection algorithm and application capable of reliably identifying specific objects within sonar data collected from standard boat-mounted sonar systems.

Brief

The client needed to detect specific objects using sonar data recorded from a boat. They had an existing software solution, however its detection performance was unreliable and inconsistent across different operating conditions.

Veluon was tasked with developing a more robust detection process and an application that could be used to analyse previously recorded sonar data.

Deliverables

We initially analysed the client's existing sonar data to understand its structure and identify patterns associated with the target objects.

We first developed detection algorithms using conventional computational techniques. These performed well under normal conditions, however detection became less reliable when objects were positioned close to the seabed.

To improve performance in these conditions, we progressed to supervised machine learning techniques. This produced significantly more reliable detection, particularly in challenging areas close to the seabed.

Veluon then developed an application that allowed the client to process previously recorded sonar data and automatically identify the target objects.

  • Analysed sonar data to identify patterns and characteristics of the target objects
  • Developed detection algorithms using conventional and machine learning techniques
  • Improved detection reliability in challenging conditions close to the seabed
  • Produced an application for processing and analysing recorded sonar data