Cameras - pros and cons

So, what makes a good camera for an autonomous camera trap for insects?

camera in a moth trap

We use a web camera in our system, which seems to work well a lot of the time, it produces a high resolution image and is fairly cheap to buy but we sometimes have issues with white balance and focus.

It seems as though the systems being developed use a range of different cameras, so I wonder if we collated out knowledge we could help to identify the best cameras on the market for our needs

I would say the key criteria are:

  • High resolution (without going overboard)
  • Cheap
  • Low power
  • Good interface for a Raspberry Pi or similar
  • Good white balance
  • Reliable auto focus

And what about processing on the edge? I was recently introduced to these cameras which look super compact, and they can do image processing (e.g. detection), on the camera itself!

In our project, we need to monitor plant-pollinator (insects) interactions in field conditions. One of the lessons I learned is that we need a camera trap capable of having a sensitive trigger for insects. Otherwise, we produce terabytes of images with no insect which comes with a lot of logistic challenges (data storage, transfer, handling, annotating, etc).

Also, to get a decent image of a pollinator so that we can identify it to the lowest taxa level possible, more important than image resolution is to have a good optical zoom or get closer to the flowers with the camera which could scare the insects.

Not sure if edge detection would work at the moment due to the higher demand on computational power (which increases easily the costs and some nano GPUs are mostly out of stock anyways) and demand on energy. However, possibly there is hardware out there to satisfy all these constraints and I would like to find out more.

Other papers that propose camera systems for monitoring plant-insect interactions:

- Droissart 2021 - PICT: A low-cost, modular, open-source camera trap system to study plant–insect  

- Bjerge, K., Mann, H. M. R., & Høye, T. T. (2022). Real‐time insect tracking and monitoring with computer vision and deep learning. Remote Sensing in Ecology and Conservation, 8(3), 315–327. 

- Ratnayake, M. N., Dyer, A. G., & Dorin, A. (2021). Tracking individual honeybees among wildflower clusters with computer vision-facilitated pollinator monitoring. Plos One, 16(2), e0239504.

- Naqvi 2022 - Camera traps are an effective tool for monitoring insect–plant interactions

Liz Ferguson
Ocean Science Analytics
Marine mammal ecologist and online technical trainer
Commenter level 1
Reactor level 1
Conversation starter level 1

I am part of a group at California State University, San Marcos using camera trap to look at insect-plant interactions. We tested out the PICT option this summer due to low cost and great instruction details. It worked quite well, jut a bit of a learning curve if you are new to Raspberry Pi. Also it became VERY challenging to find a second set of supplies due to supply chain issues, so perhaps more of an off the shelf option is better. Did the job though, and we just came across VIAME for annotating/processing video data (as the original software suggestion in the PICT documentation was not viable). 

Love to hear more about how you all evaluate data as we are new to this topic!