Rugged, High-Frequency Time-Lapse Cameras to Quantify Salmon Migrations

Professor Andrew Berdahl's lab uses computer vision to count salmon for management and to quantify salmon movement to learn more about their behavioral ecology. In principle a camera filming a small salmon stream should be able to provide a count of each salmon that entered and exits the stream along with the date and time. These data would give Professor Berdhal's lab not only highly accurate and cost effective counts but also information on how salmon may be using social information to make more accurate decisions about when and where to migrate. Professor Berdhal has been experimenting with off-the-shelf (e.g., Brinno) time-lapse cameras, but these have several fundamental drawbacks, including too slow frame rate, too low memory capacity too short battery life, and an inability to film at night (when many of the fish migrate). This student team will work to create a custom system that would address all of these issues, which would improve salmon management and contribute to the lab's understanding of how these fish make decisions. This student team will work to create a base design that is a rugged and weatherproof camera system, able to capture continuous imagery (~4 images per second) with a battery and hard drive able to last weeks-months. The student team will work to include a polarizing filter to cut surface reflection and a flash/continuous light source for nighttime images. Upgrades the student team will work to incorporate include an integrated app that: i) streams what the camera is 'seeing' to aid in set up; ii) allows the user to change camera parameters (eg frame rate); and possibly iii) wireless downloading or at least previewing data. Further expansions the student team will work to achieve include a ML system to automatically count fish passing from the captured images (the lab has made some progress on this already). The base deliverable this student team will work to achieve is a prototype of the system described above, but ideally ~10 units.

Related News