About This Project

Camera traps photos tell you what animals are in an area, but figuring out how many is tricky. E.g. two foxes in a photo doesn't mean there are only two; or a camera might see the same fox twice. But, knowing how many is key to answering questions like "Is the population of a species declining?" We're making a low-cost 3D vision add-on for camera traps. This is because 3D photos include spatial information, and existing mathematical models can use it to make much-needed population estimates.

Ask the Scientists

What is the context of this research?

Camera traps are one of the most important tools for ecologists and wildlife researchers. They are motion-triggered cameras – usually with night vision – which capture photos of wildlife in an unobtrusive way. Scientists set up a grid of dozens of camera traps in a study area to catalogue its biodiversity. However, camera traps only tell you which species of animals are present, not how many.

Measuring population sizes is a key research aim in ecological science. There are mathematical methods like the random encounter model or distance sampling which can estimate populations. But they all require a key piece of data – the distance of an animal from the observer – that camera traps do not provide. Right now, loads of manual labor is needed to obtain this critical spatial information.

What is the significance of this project?

If you search online, you'll see that there are clip-on stereoscopic lenses for smartphones, like Kúla Bebe or Poppy. They let a camera capture two side-by-side images, which – just like how human eyes give us binocular vision – can be processed into a 3D image. With the principle of distance measure by parallax, we can calculate the distance of any object in a 3D image.

Camera traps are expensive & sturdy. They need to withstand weather and animals bumping into them. So, instead of modifying camera traps, we plan to develop "3D glasses" for them in the form of stereoscopic lenses. Smartphone clip-on lenses are more expensive, and do not fit camera traps. So we will create ones that do and cost much less for cash-strapped scientists. No need to buy new cameras!

What are the goals of the project?

The main research question is if images from camera traps with 3D glasses are enough to get the spatial data needed to estimate wildlife populations.

We'll test this with camera traps of different models/brands, wearing our 3D glasses to capture images in the field. The plan is to set up "scenes", initially with fixed objects at known distances from the camera traps. The raw data will be the 3D photos of these objects, which can be processed with OpenCV to calculate their distances. By comparing these estimates with the real distances we already know, we will know how accurately our 3D glasses can obtain useful spatial data.

Once that's done, we'll also publish a data workflow that processes the 3D images into spatial data to feed into existing models to estimate wildlife populations.