Density estimation
Agouti implements tools to facilitate density estimation with models such as the Random Encounter Model by Rowcliffe et al. This functionality is available and in active development.
Agouti itself does not perform density estimation. Instead, it calculates movement and position variables that can be exported and analysed using external software such as the camtrapDensity R package.
For more information on REM analyses, see:
https://wildlifeobservatory.org/3rd-course-running-rem-analysis-on-camera-trap-data-packages-from-agouti/ https://github.com/MarcusRowcliffe/camtrapDensity
What can Agouti do?
Agouti calculates the following variables required for REM analyses:
Animal speed (m/s)
Distance of the animal from the camera (radius)
Angle of the animal relative to the camera
These variables can be exported from Agouti and subsequently used in external REM analyses.
Workflow overview
To obtain distance, angle, and speed variables for REM analysis in Agouti, all three steps below are required. If any step is missing or incomplete, Agouti is not able to calculate the density variables.
Camera calibration
Deployment calibration
Movement digitisation
Each step is described below.
1. Camera calibration
Camera calibration establishes the relationship between the size of an object in an image (measured in pixels) and its distance from the camera.
Because different camera models and image resolutions have different optical properties, camera calibration is required for every unique combination of:
Camera make
Camera model
Image resolution
Camera calibration is not required for every individual camera. If you use multiple cameras of the same model and resolution, a single camera calibration can be used for all of them.
Agouti already contains camera calibrations for many commonly used camera models. If a camera calibration is missing, a warning message will appear when attempting a deployment calibration.
In that case, camera calibration images need to be collected using a calibration pole. Detailed instructions can be found here.
The calibration images should be sent to the Agouti team, after which the camera calibration can be added to the system.
2. Deployment calibration
While camera calibration is performed once for a camera model, deployment calibration must be performed for each deployment used for density estimation.
Deployment calibration creates a two-dimensional representation of the ground surface visible in the camera image. This allows Agouti to translate pixel positions in the image into real-world locations on the ground surface.
Collecting deployment calibration images
After placing and securing the camera, you need to perform a deployment calibration using a calibration pole. Aim to collect calibration images from at least 25 different pole positions spread across the camera’s field of view. A larger number of well-distributed calibration positions will generally improve calibration accuracy.
For each calibration position, ensure that:
The calibration pole is touching the ground
The calibration pole is standing upright
If the terrain is sloped, calibration positions may follow the slope. However, avoid placing the pole on isolated raised or lowered areas that do not represent the general ground surface.
Important note: Before performing the deployment calibration, ensure that the camera will not be moved for the remainder of the deployment. If the camera position changes after the deployment calibration has been completed, any images captured after the movement can no longer be used for density estimation.
Digitising deployment calibration images
After uploading images into Agouti, each calibration pole position must be digitised.
For every deployment:
Select
Deployment calibrationin the annotation menu for the sequence containing the deployment calibration images.Place a positioning point on the highest visible marking on the calibration pole.
Place a positioning point on the lowest visible marking on the calibration pole.
Enter the corresponding heights on the pole for both positioning points.
These points allow Agouti to determine:
The visible length of the pole in pixels
The position of the ground surface
The distance between the calibration pole and the camera
Combining the information gathered from all deployment calibration positions allows Agouti to reconstruct the ground surface visible in the image by fitting a three-dimensional ground plane.
Checking calibration quality
After digitising the calibration images, use the Preview Plots button.
Two diagnostic plots are displayed:
A plot showing the digitised calibration poles within the image space.
A plot showing the relationship between image position and distance from the camera.
These plots can be used to identify potential digitisation mistakes, such as:
Incorrect point placement
Incorrect pole heights
Outlying calibration positions
In general, calibration points should align closely with the fitted curves.
3. Movement digitisation
The final step is digitising animal movement.
Movement digitisation is performed by editing an observation and placing positioning marks on the animal’s tracks as it moves through a sequence.
Creating a track
To create a track:
Open an observation for editing.
Place positioning marks on the animal’s track throughout the images in the sequence.
Follow either:
The front left foot, or
The front right foot
Positioning marks should be placed where the foot touches the ground.
Use the same foot throughout the entire track, as switching between feet can result in inaccurate speed estimates.
If the foot itself is not clearly visible, an estimated position can be used if it can be reliably inferred from the visible part of the leg.
Minimum requirements
Not every image requires a positioning mark.
However, at least two positioning marks across multiple images are required for Agouti to calculate speed.
Multiple animals
If multiple animals are present in a sequence, each individual should be tracked separately.
Each observation in Agouti can only have a single movement track. Therefore, when an observation contains multiple animals, each animal should be entered as a separate observation and tracked independently.
For example:
Instead of creating:
One observation containing three deer
Create:
Three observations containing one deer each
This allows each individual animal to receive its own movement track and speed estimate.
Calculating speed
Once a track has been created:
Keep the observation open in edit mode.
Click
Calculate Speed.
Agouti will calculate the speed for that individual track.
Exported variables
After all three steps have been completed and a new export has been created, additional fields become available in the export.
Image-level variables
individualPositionAngle: The angle of the positioning mark relative to the camera.
individualPositionRadius: The distance of the positioning mark from the camera.
Observation-level variables
Speed: The speed of the tracked animal, calculated from all positioning marks within the track and the image timestamps.