Animal size study - changing perspective with distance from camera

Overview

A series of camera trap images captured in 2010 shows a number of species - including a person - photographed from a single position. The apparent sizes of the animals and the person seem to be greatly influenced by their distance from the camera. This series of photographs is presented here, together with a composite that shows all animals, and the person, in a single frame.

Photographs

Composite image

The composite below presents all species merged into a single frame.

Composite image showing (from foreground to background), Black Currawong; Feral Cat; Spotted-tailed Quoll; Black Currawong; human walker. Courtesy, and copyright: Lyle Blackbourne.

Individual images

The individual images are presented below and may be expanded by clicking on them from a desktop browser. Individual images copyright: Chris Rehberg / Where Light Meets Dark.

Discussion

For this set of images we don't have any objective sense of the size of any of the figures, except perhaps the Currawong - but having a sense of the size of that bird depends on having had first-hand experience observing them. Feral Cats notoriously vary in size, Spotted-tailed Quolls may vary in size based on age also and nothing is known of the height of the walker. Any inference from this series of images must be related to the relative size of figures within them, and perhaps from a sense of the size of the leaves in the foreground. Despite this, a few points can be noted:

Firstly, the distance from camera to the far side of the clearing is only a few metres - perhaps 5 at most (noted from recollection - not from the image!). The walker is close to the far side of the clearing and their stride ought to be approximately 80 - 100cm - it is fairly reasonable to imagine it would only take 3 or 4 strides to walk from their position to the camera based on what can be seen of their stride and therefore the distance across this clearing is less than 5 metres.

Secondly, the foreground Currawong and Cat are probably less than 1 metre apart - judging by the leaf litter between them. They are also probably fairly close to the camera. Despite being so relatively close together, the bird appears significantly larger than the Cat - which is not what you would typically expect if the two were side by side (unless this is a very small Cat).

Third, there is an enormous difference in size between the foreground and background Currawongs - and this is despite being positioned perhaps only 4 metres apart.

It stands to reason that as an object gets nearer to the camera, it will more extensively fill the photo frame, but I think what's notable here is that even over a distance of just 4 metres, the effect on an animal's apparent size is very marked. Many camera trap images capture animals at a distances far greater than 4 metres. Consequently, such animals normally appear quite small in the frame.

Conclusions

The key conclusion to be drawn is that:

An animal appearing small in the frame of a camera trap image is not necessarily a small animal.
The apparent size of an animal is greatly reduced at even short distances from the camera.

Camera traps were originally designed for hunters that are normally targeting large game like deer, pigs and so on. The design of most commercial camera traps is that they use passive infrared technology (PIR) to detect horizontal motion within the field of view. This means an animal has to travel partway across the field of view before the camera is triggered. Additionally, and especially early during camera trap development (as with this model, manufactured in 2009) and with cheaper camera models, there is a significant delay between the camera being triggered and the first shot being captured. For this camera, which was a Scout Guard SG550 or SG560 (branded "Keep Guard" in software, as seen in the time stamp within each image), the delay quoted by manufacturers was 1.2 seconds. Many game animals will move a considerable distance in this time if they are commuting from one place to another, such as along a trail.

For all of these reasons, most camera traps have a wide angle of view of the scene before them - this allows for the fact that during the 1.2+ seconds before the images is captured, the animal may have traveled partway across the frame. The wider the camera's field of view, the more likely the animal will still be in-frame when the photograph is finally taken. The trade-off, however, is that the wider the field of view, the smaller the animal will appear - and this is accentuated when the animal is also some distance from the camera. This may be fine when hunting large deer, but can make identifying and measuring smaller animals quite difficult.

(Some camera models do work to reverse this trend - providing a narrower field of view so that the animal appears larger in frame, even when distant, but the tradeoff in this scenario is that there is a greater chance the animal has already moved past the field of view by the time the photograph is taken. Such models work well when monitoring wildlife that stays put - such as with birds at a nest.)

Put another way, the above conclusion implies:

The apparent size of an animal is greatly influenced by its distance from the camera when using camera traps with wide fields of view, and
It may be quite difficult, if at all possible, to estimate an animal's size unless rigorous steps are taken to provide scales within the images.

All of these details combined suggest it is probably quite difficult to produce an accurate estimate of an animal's size based on camera trap photographs. Even when a painted stake (with markings at known intervals) appears in frame, unless the animal is exactly the same distance from the camera as the stake, the animal will not be measurable at the same scale as the stake - and the difference between the two will be unknown.

Calculating animal sizes in camera trap images

For the reasons given above, using a single marker within a camera trap photo will prove ineffective in helping estimate accurate animal sizes in many cases.

The photograph below comes from a camera trap review and shows several dummy deer standing in a North American forest, each with a marker showing how many feet the animal is from the camera.

Image captured during testing of a camera trap. Copyright: Chasingame.

In this case, nothing is known (by this author, at any rate) of the size of the deer - but if the deer were replaced with stakes that had been painted with marks at fixed intervals (say, 10 centimetre intervals), then the photograph would be capturing 6 such stakes: one at 10 feet (note the "10" marker in the foreground); one at 20 feet, and so on. As it stands, it seems probably that the distance markers in this photograph (showing "10", "20", and so on) are probably all of uniform height - and could be used as a de facto scale by measuring their real-world height.

As long as the animal's feet can be seen - so that its distance from camera may be estimated - a researcher could establish the approximate distance from the camera (for example, something like "between 30 and 40 feet"). Further, the stakes placed just in front of, and just behind the animal, can be used as scales within the photograph. That is, we could easily calculate that "at 30 feet, X number of pixels correlates to 10 centimetres" - or whatever interval is appropriate (and marked on the stake). A similar calculation could be made for the second stake. In this way a size range for the animal could be generated, along the lines of saying "if the animal was at 20 feet, then it would be X centimetres tall; but if it was at 30 feet, it would by Y centimetres tall".

The animal would, of course, be between 20 and 30 feet from the camera (in this hypothetical example) and therefore its actual size would be somewhere between the two estimates. This could be refined further if a fair assessment could be made as to whether the animal is nearer the foreground or background stake.

One final consideration is that because of the camera's wide field of view, a stake placed some distance - say X - directly in front of the camera may appear further away in the frame (positioned higher, vertically) than one placed exactly the same distance, but at the edge of frame. There would be merit in using multiple stakes in this case - both at the edge of frame and at the centre.

Clearly it can be seen that the effort required for an accurate estimate of animal size based on camera trapping is quite considerable. This may preclude the above approach from being used for every camera deployment, but an alternative approach might be to ensure that when a camera is deployed, its exact position is well and truly noted and/or well and truly fixed (such as to an immovable tree) - so that if an animal of interest is captured, exactly the same camera may be re-deployed in exactly the same position for the explicit purpose of setting up and photographing marker stakes as just described. The advantages here are that there is no overhead in setting up stakes for every camera deployed, there is no risk of stakes falling over or being knocked over by wildlife and there is no need to have a large number of stakes available where multiple cameras are deployed simultaneously. Preparing size estimates will involve only a little more complexity - in ensuring the camera's field of view when photographing the stakes is as near to identical as the original deployment, and then in merging two photographs: the photograph of the animal being measured, with the photograph of the stakes.

References

Chasingame, 2017. "2017 Primos Proof Cam 01 Camera Review" accessed 25 Jan 2021 at https://www.chasingame.com/2017-primos-proof-cam-01-camera-review/.