Login / Signup

A non-invasive approach to measuring body dimensions of wildlife with camera traps: A felid field trial.

Alexandra J PatonBarry W BrookJessie C Buettel
Published in: Ecology and evolution (2024)
Dimensions of body size are an important measurement in animal ecology, although they can be difficult to obtain due to the effort and cost associated with the invasive nature of these measurements. We avoid these limitations by using camera trap images to derive dimensions of animal size. To obtain measurements of object dimensions using this method, the size of the object in pixels, the focal length of the camera, and the distance to that object must be known. We describe a novel approach of obtaining the distance to the object through the creation of a portable distance marker, which, when photographed, creates a "reference image" to determine the position of the animal within an image. This method allows for the retrospective analysis of existing datasets and eliminates the need for permanent in-field distance markers. We tested the accuracy of this methodology under controlled conditions with objects of known size resembling Felis catus , our study species, validating the legitimacy of our method of size estimation. We then apply our method to measure feral cat body size using images collected in Tasmania, Australia. The precision of our methodology was evaluated by comparing size estimates across individual cats, revealing consistent and reliable results. The average height (front paw to shoulder) of the feral cats sampled was 25.25 cm (CI = 24.4, 26.1) and the average length (base of tail to nose) was 47.48 cm (CI = 46.0, 48.9), suggesting wild feral cats in our study area are no larger than their domestic counterparts. Given the success of its application within our study, we call for further trails with this method across a variety of species.
Keyphrases
  • deep learning
  • convolutional neural network
  • working memory
  • randomized controlled trial
  • clinical trial
  • body mass index
  • optical coherence tomography
  • physical activity
  • study protocol
  • machine learning