Understanding Biological Motion using Moving Light Displays

使用移动光显示器了解生物运动

• 批准号: EP/D506549/1
• 负责人: Neill Campbell
• 金额: $ 45.21万
• 依托单位: University of Bristol
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2006
• 资助国家: 英国
• 起止时间: 2006 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=EP%2FD506549%2F1
• 关键词: Understanding; Biological; Motion; using; Moving

The major inspiration for this work is that the automatic extraction and understanding of biological motions is largely an unsolved problem. Because of the complexity and variability of natural motions (and hence the image sequences that encode them), automatic computer vision based analysis of such data is under-developed. Current techniques rely on highly constrained data collection methods or intrusive physical interaction. Traditional motion capture relies on the placement of trackable locators (often affixed to lycra suits) which is often impractical and sometimes impossible. Apart from the expense and the need for a constrained filming environment, locators cannot easily be placed on wild or small creatures. To be able to automatically extract characteristic motion from existing film footage is desirable since many applications could benefit from such information. Such applications include media database searching and understanding, accurate automated surveillance systems, medical based gait analysis, computer animation and robotics.We propose that a computer vision based approach to the understanding of the Moving Light Display phenomenon will improve, in a general sense, the interpretation of motions shown in image sequences of animated animals and humans. The MLD (also known as the Point Light Display, PLD), consists of a small set of point lights that describe an action or behaviour (walking or dancing for example), of a subject. A great deal of work has been carried out on the analysis of MLDs and the design of stimuli for psychophysical experiments. Psychologists have shown that high-level information can be extracted from the temporally varying point light set moving such that actions or behaviour such as gait, personal identity, gender and even mood can be determined. It has also been shown that an average observer can deal with a significant amount of MLD perturbation such as noise, occlusion, translation and rotation (in both 2D and 3D). This observation implies that a large amount of information can be extracted from a small set of moving point lights.It is interesting to note that all the major advances in human tracking utilising computer vision over the last ten years, have been based around the appearance (colour, texture etc.) of the subject. This is not surprising since, using a constrained or known background and making assumptions about coherent clothing, strong edges or skin colour, such approaches have led to some impressive algorithms and systems. One of the main claims here, however, is that by doing so, the field at large has missed an important step. A small number of point lights representing an action, motion or emotion is easily perceived by humans, but we still have almost no underlying theorems as to why this should be so. Crucially, we have no 'manual' as how to transfer this skill of humans so as to provide an engineering solution to a range of important visual inspection tasks. This proposal aims to address this gap in our knowledge.

这项工作的主要灵感是，生物运动的自动提取和理解在很大程度上是一个未解决的问题。由于自然运动的复杂性和可变性（以及因此编码它们的图像序列），对这种数据的基于自动计算机视觉的分析还没有得到充分开发。目前的技术依赖于高度约束的数据收集方法或侵入式物理交互。传统的动作捕捉依赖于可跟踪定位器的放置（通常固定在莱卡套装上），这通常是不切实际的，有时甚至是不可能的。除了费用和拍摄环境的限制，定位器不能轻易地放置在野生或小型生物身上。能够从现有的电影胶片中自动提取特征运动是期望的，因为许多应用可以受益于这样的信息。这些应用包括媒体数据库搜索和理解，准确的自动监控系统，基于医学的步态分析，计算机动画和robotics.We建议，基于计算机视觉的方法来理解的运动光显示现象将提高，在一般意义上，解释动画动物和人类的图像序列中所示的运动。MLD（也称为点光源显示器，PLD）由一小组点光源组成，用于描述对象的动作或行为（例如行走或跳舞）。大量的工作已经进行了MLD的分析和心理物理实验的刺激设计。心理学家已经表明，可以从移动的时间变化的点光集合中提取高级信息，使得可以确定诸如步态、个人身份、性别甚至情绪的动作或行为。它还表明，一个平均的观察者可以处理大量的MLD扰动，如噪声，遮挡，平移和旋转（在2D和3D）。这一观察意味着大量的信息可以从一个小的移动点光集合中提取出来。有趣的是，在过去的十年里，利用计算机视觉进行人类跟踪的所有主要进展都是基于外观（颜色，纹理等）。的主题。这并不奇怪，因为使用受约束或已知的背景，并对连贯的服装，强边缘或肤色进行假设，这些方法已经产生了一些令人印象深刻的算法和系统。然而，这里的一个主要主张是，这样做，整个领域错过了重要的一步。少量的点光源代表一个动作，运动或情绪很容易被人类感知，但我们仍然几乎没有基本的定理来解释为什么会这样。至关重要的是，我们没有“手册”来说明如何转移人类的这种技能，以便为一系列重要的目视检查任务提供工程解决方案。这项建议旨在填补我们知识上的这一空白。

项目成果

Form overshadows 'opponent motion' information in processing of biological motion from point light walker stimuli.

在处理来自点光步行者刺激的生物运动时，形式掩盖了“对手运动”信息。

• DOI: 10.1016/j.visres.2009.10.021
• 发表时间: 2010
• 期刊: Vision research
• 影响因子: 1.8
• 作者: Thirkettle M