Skip to main content
Dryad logo

A motion-detection based camera trap for small nocturnal mammals with low latency and high signal-to-noise ratio

Citation

Klemens, Jeffrey; Tripepi, Manuela; McFoy, Shane (2021), A motion-detection based camera trap for small nocturnal mammals with low latency and high signal-to-noise ratio, Dryad, Dataset, https://doi.org/10.5061/dryad.m0cfxpp3m

Abstract

1. Camera traps are useful for monitoring wildlife populations, but traps may not always trigger when targeting small, nocturnal species. Motion detection techniques have advantages over time-lapse and heat-triggered traps, but need to be deployed to maximize signal-to-noise ratio. 2. As part of a study of flying squirrels (Glaucomys) in urban environments we developed motion detecting camera traps using a raspberry pi microcomputer and camera and a 940 nm IR illuminator on a tree-mounted wooden platform. The system was built from commercially available parts and was comparable in cost to a consumer camera trap, although this cost did not include a waterproof housing. We compared the performance of our system to commercial trailcams. 3. Four pi cameras successfully documented visits by Glaucomys and other animals to bait placed on the platform over three nights at four wooded sites: suburban and rural backyards, a private outdoors club, and a small urban nature reserve (16 camera X site combinations, 48 trap nights). The traps showed low latency, with an average of < 1 night until detection of Glaucomys at each site. Data collected had a high signal-to-noise ratio; of 2182 capture events 55% documented Glaucomys, 40% documented non-target mammals, 1% were caused by large insects, and the remaining 4% were unknown. Commercial camera traps placed at the sites failed to capture many of these events. 4. The low cost and high signal-to-noise ratio of this system may make it easy to adapt for other small animal applications. The main modifications required to deploy this system in new situations will be in locating or providing a fixed or static background against which animals can be observed and in using masking techniques within motion detection software.

Methods

This dataset describes trailcam images generated from four Raspberry Pi-based trailcam prototypes and associated commercial trailcams as described in the associated MS (Klemens, Tripepi, and McFoy 2021). The data was generated by reviewing all motion detection events generated by the cameras and attributing them to either flying squirrels (Glaucomys sp.), other mammals, insects, or 'other', which includes capture events that cannot be attributed to an organism. The data records the total number of capture events recorded between placement of bait at the camera trap and sunrise, and then attributes each events to one of the four categories.

For some but not all site by camera combinations commercial camera traps were used to observe the Pi cams. The insect category is excluded as trail cams did not create events due to insect movement. The number of capture events in each of the other categories is presented. Percentages presented for the trailcam data represent the percent of capture events recorded by the Pi camera that were also captured by the trailcam.    

Location data was collected using a Garmin GPSMAP64ST and the default WGS 84 map datum and WGS 84 map spheroid.

Usage Notes

Data.csv contains the records for each camera trap for each night of the experiment. 

Locations.csv contains the lat/long data for each camera trap X site combination