HomeBlogAvoiding False Triggers in Trail Cameras

Comments

Avoiding False Triggers in Trail Cameras — 12 Comments

  1. Great article. Personally when I install a new trail camera in a place with potential high false triggers, I try to set it in a conservative way, this is with long intervals (like 25 sec), photo mode, low megapixels. Then I check it a couple of weeks later, if it’s all good then I usually change it to video or a better quality photo with low interval (like 5s). After that, I can let it for several months with confidence. I mostly use trail cameras for research so the camera has to work at all times, loosing trail camera time due to battery or sd card storage has to be avoided at all cost. BTW, here in the southern hemisphere we have to place the camera facing south. Cheers from Chile.

    • “Trying out” a set certainly makes a lot of sense. And thanks for confirming the location of the sun for our southern hemisphere readers 🙂 (I suspected this was true, but couldn’t convince myself without a picture). I will add to the post. Thanks for reading and commenting!

  2. Great stuff. Thank you. We are out in the Western US and the wind is a steady companion on the open prairie. There are indeed such places where there is nothing one can do to avoid those false triggers and yes, they are usually in the middle of the day. Almost like clockwork.

    • Glad you found it useful. For “clockwork” triggers, have you tried reducing operating hours? Or, like us, do you always find yourself hoping to find something really good in the haystack of false triggers?

  3. Thanks for this article. I thought PIR sensors primarily picked up movement, not realizing the importance of temperature change and its effect on the PIR sensor. With respect to your suggestion to cut off the tree just above where the camera is attached makes complete sense; however be mindful that the camera and lock may be stolen simply by pulling them off the tree! I had locked one of my cameras on a tree with about a 2.5 inch diameter. Someone bent the tree to break it, and slipped my camera and lock off.

    Something I’m very curious about is why moving water doesn’t trigger a camera, even when the camera is set close to the water, yet when the camera is set close to moving vegetation it’s triggered incessantly. Is this because there aren’t temperature changes with moving water, but there are with moving vegetation? And, if this is the case, I wonder why there aren’t temperature changes with moving water….

    • Sorry to hear about you camera 🙁 You’re absolutely right about the danger of mounting cameras to small trees. Especially if you cut them to keep them from blowing around 🙂 These days, we deploy our cameras only in steel lockboxes. We have also taken to screwing these into the trees through the mounting holes in the back. A cable lock through the side holes and/or a padlock through in front secures the camera (and prevents access to the screws). This makes it much more difficult for a thief to remove the lock box or camera. It also makes it easier to check the camera without requiring us to re-aim the set. Since the lock box is screwed in and doesn’t move, we just pop the camera out and then back in. That being said, nothing works 100% against theft.

      Water is tricky because it has lots of interesting behaviors. It flows, obviously, and yet relatively high thermal conductivity means all of the water in a stream passing by your camera is likely at the same temperature, so there is no “moving changes in temperature” to trigger the camera. On the other hand, it can also lap up against non-water edges, possibly at different temperatures, leading to changes in temperatures as water covers, then recedes, which could cause triggers. Finally, water is a good reflector of the Near IR radiation detected by the PIR sensor. The surface of flowing, or blowing water, can therefore act as a rippling mirror, causing a static thermal background to appear to the PIR sensor as “moving” temperature — another source of false triggers. Having said all this, your question begs some “hard data”. I’m going to take my thermal imaging camera out to our next water set, and collect some data. Perhaps this will make it into a future post just on water!

      Thanks for reading and commenting.

      • Thank you very much! I would be very interested in seeing a post just on water. Clearly, it’s a very interesting element when it comes to trail cameras. I currently have a camera set up by a flowing river, just off shore (hoping to get grizzlies feeding on salmon). Last year, when I set up a camera along the river, I was surprised and pleased to not have gotten any false triggers. The current (no pun intended) setup has been there for about a month now and I’m hoping to get back to it within the next few weeks, however, I’m dependent on someone taking me in their boat. Nevertheless, when I get the camera, I’ll let you know if I get any false triggers and also if I don’t! Thanks again, Mary Beth

  4. Hi all
    I think with excessive false triggering there’s another problem/risk that arises: In a situation when the cam captures videos nearby constantly (seamlessly) because of false tiggering, any animal activity during this period of time will be recorded on a video that’s already running. There’s no triggering of the cam – the cam is already running. This means that the animal activity is most often not visible within the first few seconds of the video containig it. As inspecting the whole duration of every single video is extremely time consuming, one tends to check only the first few seconds of each video for animal activity, and to delete any video that shows no action within this short period of time. So even recorded animal activities will get lost. Hmmm… I wonder wheter anybody knows a program/software beeing able to run a batch check running through a large heap (hundreds) of videos for any unusual (animal) movements, and beeing able to create a list of videos containing such an “anomaly”. Having something beeing able to do this would massively reduce the risk to loose precious captured animal movements, and it would save all of us a lot of time we can spend for better things by scanning through mountains of useless videos.
    What do you think? Please let my know, I’m curious 🙂

    • You are right! We have often had the experience of thinking there might be camera trapping gold hiding among gigabytes of false triggers. More generally, it would be a great if there were a way to automatically find all videos with animals, perhaps sorted by species and “quality”. I believe the technology exists for such a system. For example:

      * MegaDetector: Is a highly accurate, open source, AI model for finding animals in trail cameras (much more subtle than finding animals in photos taken by people). https://github.com/microsoft/CameraTraps/blob/main/megadetector.md While MegaDetector doesn’t give a “quality” metric, it does provide a bounding box for each animal detected, and a probability that it’s identification is right, both of which could be used to inform a crude “quality” metric.

      * MegaDetector runs a frame at a time. This is fine for still photos. For videos, one would have to split the MP4’s into a sequence of frames, and process each frame through MegaDetector. I am aware of some experimental models that take information from multiple frames (the way we do when we flash back and forth between frames) to improve predictions, but not one that operates on seconds worth of video.

      * MegaDetector is designed for accuracy, and it’s very good, but it requires a fairly beefy server running with a GPU or two. I’ve had luck running it on the free version of Google Colab, though it runs even better on the (USD $10/month) “professional” version.

      * The trick is then to get the data from the SD card out to the cloud where Megadetector running on Google Colab GPU nodes can process it. Unfortunately, with typical upload speeds, it takes a long time to do the easy thing, which is to transfer 10’s of GBytes/SD card up where it can be processed. A useful system could be built by taking a few shortcuts, namely: reducing the resolution of each image to the native resolution of MegaDetector; and processing every Nth frame from the video (instead of all frames). This would all have to be done locally, but it would substantially reduce the amount of data that had to be shipped up into the cloud

      I’ve prototyped this whole sequence, and it all works, in principle. The prototype is very slow (does not include all the shortcuts, above. But it’s a straightforward integration of existing tools. But I’ve been working on other things, and haven’t gotten around to packaging it up.

      -bob

  5. While trying to capture images of a raccoon on a garage roof using a trail camera, the only images I was able to capture occurred in the early morning and they were ghost-like. This I attribute to the high background temperature of the roof. I might have had much better results on a cold day.

    • What do you mean by “ghost like” ? (sorry that WordPress doesn’t allow photo attachments in comments 🙁 ) Black and white? Underexposed? Overexposed? Mirage like? Is the roof really that hot in early morning? Let me know if you’d like me to look at the photo, and I’ll contact you by email for a copy. Or let me know if it’s available somewhere online (e.g. FB)

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

HTML tags allowed in your comment: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

Shares