Monitoring fish species to keep track of their population growth and general health is vitally important. Current techniques for assessing fishes range from “capture and release” methods to monitoring fish passively with imaging devices. Not surprisingly many custom imaging systems have been built for underwater imaging. One-off stereo camera systems, like that seen in the image below, are common, yet difficult to use not only because of the noise and inaccuracies inherent in the system, but also because of the bulkiness and lack of real-time feedback.

FishSense aims to take advantage of commercial depth cameras, which have been making tremendous advances in recent years related to imaging quality, performance, and cost. Our initial work centers around the Intel RealSense D455 camera, which combines infrared depth imaging and traditional RGB color imaging to construct a 3D picture in real time. Together, with collaborators from Scripps Institute of Oceanography, we hope to validate the effectiveness of this technology and open up this field to high resolution, short range 3D imaging.

We are currently developing a prototype handheld depth imaging platform that will enable amateur citizen science divers to collect professional grade data. We aim to integrate the RealSense D455 into the VIAME toolkit to leverage their existing analysis techniques, but also build support for Intel RealSense cameras.

• 

• 

If you wish to learn more about this project, contact Chris Crutchfield (ccrutchf@eng.ucsd.edu).