Markerless motion capture applied to the analysis of locomotor kinematics in the semi-aquatic hunting spider, Dolomedes aquaticus

Abstract

This thesis focuses on two key goals. Firstly, developing a markerless motion capture technique for the examination of joint angles during spider locomotion. Secondly, applying this technique to understanding gait and gait generating mechanisms in the semi-aquatic hunting spider, Dolomedes aquaticus.

I present a markerless technique for reconstructing 2D joint angles during locomotion based on information contained in video frames and a spider model based on the relative lengths of segments and joint angle limits. This algorithm allows gait analysis without the need for a sophisticated lab setup. Analysis is based on the subject filmed by a stationary video camera. Techniques that recover body pose from video sequences with little user intervention have numerous applications such as motion capture, gesture recognition, surveillance of people or animals and animation for movies or computer games. The spiders’ pose is estimated in every frame of a video sequence. The basic elements of my tracking approach include an articulated body model, extracted features from video frames and various constraints. These components are combined in a Bayesian framework, which segments the frame into foreground and subject and estimates the pose of the subject. Joint angles are used to investigate gait and gait generating mechanisms underlying locomotion in the spider. Firstly, kinematic parameters were compared to mass and body length of spiders. Stride length was the only kinematic parameter to yield significant results compared to spider size, however non-significant scaling relationships were similar in magnitude to those in the literature. Secondly, kinematic parameters were analysed in relation to speed of locomotion. Stride frequency showed a greater correlation with normalized speed than absolute speed and stride length showed a greater correlation with absolute speed than normalized speed. This suggests that larger animals increase their speed by increasing stride length, whereas it is possible for smaller animals to increase their speed by increasing stride frequency. Contrary to the relationship frequently observed in insects, both protraction and retraction periods decreased with speed. Thirdly, changes in velocity and acceleration were compared across the trajectory of each pair of legs and the ipsilateral and contralateral coordination of legs was investigated. Each leg was found to contribute in a specific manner to locomotion. Movements of front legs were random, suggesting they play some other role, possibly sensory, rather than contributing to stability. Legs 2 and 3 appeared to play a more dominant role in generating propulsive force, with hind legs probably contributing more to stability than propulsion.