Markerless Motion Capture for Primate Locomotion Studies

人类的移动方式是独一无二的，因为我们是唯一能够用完全伸直的腿走路的动物。我们认为，这一特征发生在人类谱系从黑猩猩谱系分裂出来后不久，因此了解我们是如何移动的将有助于我们理解这种分裂是如何发生的。为了了解人类行走和奔跑的进化，我们还需要了解我们的近亲是如何行走和奔跑的，这导致了许多研究人员研究非人类灵长类动物的运动。最常见的方法是用摄像机拍摄动物的所作所为。然而，在实践中，这项技术存在许多困难。从录像带上进行测量，要么需要我们在研究动物身上放置反光标记，这通常是不可能的，要么需要由熟练的研究人员单独测量各个帧，估计灵长类动物的关节在哪里。这不是很准确，因为通常关节位置会被头发遮挡。我们经常想要了解3D模式下发生了什么，因为灵长类动物花了很多时间在树上，它们的动作非常复杂，这特别难通过简单的视频分析来量化。幸运的是，在过去的几年里，计算机技术已经发展到了这样的程度，只要它们重叠，我们就可以从一系列照片中重建出三维形状。这种立体重建的工作原理是识别共同的特征，就像我们眼睛中立体视觉的工作方式一样。因此，对于高清晰度视频，可以在3D中提取被摄体动物的皮肤轮廓。这只是战斗的一半，因为我们真正想知道的是骨骼在皮肤下是如何移动的。幸运的是，我们也有从CT扫描中获得的骨骼的准确3D模型，可以放在皮肤包膜中，这应该会给我们所需的信息。这个项目的目的是测试这一新方法的有效性。我们可以将使用新技术获得的结果与使用一系列替代标准技术获得的结果进行比较。通过使用受控条件下经过训练的猴子作为我们的研究动物，我们甚至可以使用非常精确的基于标记的系统作为‘金标准’参考，以确定新技术的局限性，并通过在一系列不同的动物身上尝试该技术，我们可以看到毛发长度或颜色等特征是否会造成任何问题。最终目标是创造一种更好的技术来研究灵长类动物的运动，这将大大减少所需的努力，同时产生更好的结果。这些改进的数据是必要的，可以让我们了解现代人行动方式进化的全部复杂性，以及我们与灵长类中最接近的亲戚有何不同。

Humans are unique in the way they move since we are the only animal that is able to walk on fully straight legs. We think that this feature occurred very soon after the human lineage split off from the chimpanzee lineage so that understanding how we came to move the way we do will help us understand how this split occurred. To understand the evolution of human walking and running we need to also know about how our nearest relatives walk and run and that has led to many researchers studying non-human primate locomotion. The commonest approach is simply to use a video camera and film what the animal is doing. However in practice there are many difficulties with this technique. To take measurements from video film either requires us to put reflective markers on the study animal which is usually simply not possible, or the individual frames need be measured individually by a skilled researcher estimating where the primate's joints are. This is not very accurate because often the joint positions are obscured by hair. Often we want to understand what is happening in 3D since primates spend a lot of time in the trees and their movements are very complex and this is particularly difficult to quantify from simple video analysis. Fortunately in the last few years computer technology has advanced to such an extent that we can reconstruct three-dimensional shape from a series of photographs as long as they overlap. This stereo reconstruction works by identifying common features in much the same way that the stereoscopic vision works in our eyes. With high definition video it is therefore possible to extract the skin outline of a subject animal in 3D. This is only half the battle since what we really want to know is how the skeleton is moving underneath the skin. Fortunately we also have accurate 3D models of the skeleton obtained from CT scans that can be fitted within the skin envelope and this should give us the information we require. The purpose of this project is to test the effectiveness of this new approach. We can compare the results we obtain using the new technique against those we obtain using a range of alternative standard techniques. By using trained monkeys in controlled conditions as our study animals we can even use the very accurate marker based systems as a 'gold-standard' reference to identify what the limitations of the new technique are, and by trying the technique on a range of different animals we can see whether features of the fur such as hair-length or colour cause any problems. The ultimate goal is to produce a much better technique for studying primate locomotion that will greatly reduce the effort required whilst at the same time producing better results. This improved data is necessary to allow us to understand the full complexity of the evolution of the way modern humans move and how we differ from our nearest relatives among the primates.