Augmenting 3D Modeling and Animation Techniques Using Large Sensed Data

三维（3D）计算机图形学一直是我的主要研究领域。尽管取得了许多进步，但创建复杂的3D几何模型和动画仍然是一项工作量很大的工作，并带来了许多挑战。如今，低成本相机和其他传感设备的广泛使用使得能够快速捕获图像、视频、3D表面和运动数据。再加上多核CPU和多核GPU的持续增长，这些技术为我们提供了巨大的机会，可以改变我们执行3D形状建模和动画任务的方式。该提案的具体研究重点是调查计算方法，有效地处理这些设备捕获的大量数据，并获得合适的表征表示。然后，这些可以用于增强3D建模和动画的技术。通过这项研究，我希望回答以下一些问题：1）什么形状特征可以从3D扫描计算用于形状重建，检索等？2)确定哪些是有助于创建/增强动画的适当运动数据特征？3)我们如何从图像/视频数据中学习，以告知3D形状和动画创建过程？4)我们如何使用来自Kinect类设备的感知数据来扩展具有全身运动的3D交互？5)我们需要什么特定的编程模型和技术来并行化以上开发的建模和动画技术在商品异构计算环境？我的研究方法基本上是由研究，分析，新想法，实施，测试和验证的循环驱动的。我的计划是建立在我的3D图形实验室正在进行的研究，该实验室具有必要的图形数据捕获基础设施。我将继续与美术研究人员的现有合作，为我们提供游戏和娱乐中的真实使用场景。这项研究的主要目标是使设计师和动画师更容易为新的情况制作3D模型和动画，而不需要过多的用户输入，编程或人工努力。

Three dimensional (3D) computer graphics has been and is my dominant area of research. In spite of manyadvances, creating complex 3D geometric models and animations continues to be effort intensive and poses many challenges. Today, the wide use of low cost cameras and other sensing devices enable rapid capture of image, video, 3D surface and motion data. Combined with continued growth in multi-core CPUs and many core GPUs, these technologies present tremendous opportunities to transform the way in which we can carry out 3D shape modeling and animation tasks. The specific research focus of this proposal is the investigation of computational methods that efficiently process the large volumes of data captured by such devices and derive suitable characterizing representations. These can then be used to augment techniques for 3D modeling and animation. Through this research I hope to answer some of the following questions: 1) What shape characteristics can be computed from 3D scans for use in shape reconstruction, retrieval, etc.? 2) Identify what are appropriate motion data characteristizations which help create/enhance animations? 3) How can we learn from image/video data to inform 3D shape and animation creation processes? 4) How can we expand 3D interaction with full body movements using sensed data from a Kinect like device? 5) What specific programming models and techniques do we need to parallelize above developed modelling and animation techniques on commodity heterogeneous computing environments? My research methodology is essentially driven by the cycle of research, analysis, new ideas, implement, test and verify. My plan is to build upon on-going research in my 3D Graphics lab which has the necessary infrastructure for graphics data capture. I will continue the existing collaboration with Fine Arts researchers to provide us with real-life usage scenarios in games and entertainment. The principal goal of this research is to make it easier for designers and animators to produce 3D models and animations for novel situations without requiring excessive user input, programming or human effort.