Scorpion Vision Gives Site To Autonomous Underwater Vehicle

Scorpion Vision Gives Site To Autonomous Underwater Vehicle

In the summer of 2009 a team of undergraduate students at the University of Southampton joined eight other University teams across Europe to compete at the QinetiQ Ocean Basin

The Delphin AUV

Tank with vehicles they had designed and built themselves at the fourth annual SAUC-E competition.
‘SAUC-E challenges the next generation of engineers to design and build an Autonomous Underwater Vehicle (AUV) and then perform realistic missions  in the underwater environment. The event is designed to encourage young engineers and scientists to think about underwater technology and its future possibilities, as well as fostering interest in innovation and technology, and encouraging careers in the field‘. – UK Defence Science and Technology Laboratory
The team at Southampton selected Scorpion Vision Software for the navigation system, utilising two low cost Unibrain firewire cameras and a TIS video converter to take video stream from a third analogue camera.
The aim of the competition was for the AUV to successfully navigate round the competition course – an underwater obstacle course laid out with gates, moving targets and a docking station. So an obstacle avoidance and navigation application was created using the in-built tool-set inside Scorpion. Using standard vision tools and some Python scripting, the team were able to create a robust guidance system with only a small amount of programming. The camera interfacing, video capture and image processing were all taken care of by Scorpion, allowing the team to focus on the other challenging aspects of the build.
Creating a navigation system using a vision system is not child’s play. Creating a navigation system that will work autonomously underwater is a major challenge! Using Scorpion for this task meant that no machine vision code needed to be written as it is already available within the Scorpion framework. Furthermore, the vision  application can run in parallel on the Windows PC along with the other mission critical applications.

As Scorpion can process the vision tools in super quick time, multi-tasking is not a problem.
The outputs from the vision system are identical for all objects and consist of:·

• A target ID specifying which target has been identified.
• A camera ID specifying which camera identified the target
• A horizontal position of the target within the image
• A vertical position of the target within the image
• An approximate size of the target which can be used to determine the relative distance between the vehicle and the object.

The team successfully built their AUV and with a high level of confidence, entered the competition. Sadly however, they were beset with technical problems from the outset as they could not get the internal PC to boot up in time for the start of the competition. However, they were still able to obtain a creditable 4th place as once the initial problems had been solved they were able to demonstrate a first class autonomous vehicle that navigated the complete course.
Further information: http://www.dstl.gov.uk/news_events/competitions/sauce/09/journals/Delphin.pdf
All components used in the navigation system are available from our on-line catalogue, the vision software is available as a free download.