BACKGROUND: Identifying how animal signals are optimized through evolution is reliant on strategies that allow us to quantify signals (and noise) in a meaningful way. Signals defined by movement presented a challenge for analysis until the application of computer vision algorithms made quantifying movement more straightforward. However, these approaches calculate image motion from two-dimensional image sequences that are restricted to a single camera view and thus under-represents the signal. The problem is exacerbated when signals have a complex three-dimensional (3D) trajectory.

  Tail flicking by the Australian Jacky lizard (Amphibolurus muricatus) is one such signal, and mediates the efficacy of signalling in this species. Tail positions are not confined to a single plane of movement, but rather move randomly in 3D space around the lizard’s body. Complicating quantitative analysis further is that the tail is not rigid during displays but constantly changes shape to resemble the flicking motion of a whip. Signals that feature a complex 3D trajectory are common, yet no attempt has been made to preserve this design feature when quantifying structure.

AN INNOVATIVE SOLUTION: We propose a powerful new strategy to address these constraints. The key components are:

• Motion capture – filming real lizard displays using cameras aligned in space and digitizing the signal - View/hide image

Left: Filming setup for pilot work showing 3 aligned cameras (although only two are required).
Right: Digitized positions of the tail tip over time in 3D space



• Animations – motion capture data recreates the signal in a virtual 3D environment - View/hide image


The digitized positions define the movement of the lizard (the tail, in this example). A texture map is then overlaid on the 3D wireframe model to give the realistic appearance of lizard skin (left). The partial tail images (right) were used in our pilot work to demonstrate the utility of working in an animation environment. Adjustments to the texture map changed the appearance of the tail causing the distal two-thirds to 'disappear'. The tail moves precisely the same way and so we can consider how tail structure and appearance influence signal efficacy with unprecedented control.




• Virtual microhabitats – recreate plant habitats in a virtual environment such that we can simulate varying microhabitat types and environmental conditions - View/hide details


Planted habitats can be reproduced in amazing detail. Plants can be individually constructed and possess unique responses to envornmental conditions. The adjacent image is a screenshot from the software we propose to use Vue8 - xStream by e-on software. Click on the image to go to their website for further information.




• Scene analysis – ‘recording’ animations under a variety of conditions and from any viewing position (angle and distance from signaller). Scenes can then be subjected to saliency analysis to identify the relative efficacy of the signal. - View/hide details



Animation sequences (whole scenes: signal and plants) can be analysed and subjected to motion analysis. Motion information can then be used as input for saliency analysis to predict the most salient motion information in the scene and therefore estimate the signal's efficacy under the prevailing conditions.

OUTCOMES: This proposal will develop a new methodology, answer recalcitrant questions in evolutionary biology and provide a roadmap for further study:

• Develop a new methodology – providing unique flexibility.
• Answer recalcitrant questions – regarding the factors influencing movement-based animal signal design/evolution.
• Provide a roadmap for further study – to target research effort of free-living Agamid lizards of Australia.

Participants: Richard Peters (La Trobe Uni) & Shaun New (The ANU)