Collaborative Proposal: ABI Innovation: Rapid, Interactive, Visual Mining of Biological Motion

在整个生物科学中，高质量、高帧速率的图像和视频被广泛用于分析亚细胞细胞器到单个生物体。视频的使用已经变得如此普遍，以至于许多研究团队积累了比目前方法合理分析的数据多得多的数据。该项目将开发算法，以允许快速和负担得起的非常大的视频集挖掘感兴趣的运动行为，在研究复杂的生物体系统的集体行为和紧急特性。这项工作是由生物运动分析各个领域的专家的需求推动的，并结合了计算机视觉、图像处理、用户界面和可视化方面的研究。该项目的一个关键部分是一种快速、半自动的生物对象跟踪器，它利用现成硬件的计算能力，并结合简单的用户交互，在生物图像分析中经常出现的具有挑战性的情况下显著提高跟踪精度。领域专家将能够通过视觉查询选择感兴趣的示例运动来搜索运动数据库。该系统将在三个生物学研究应用上进行评估：蜜蜂行为、蚁群网络和细胞运动分析。生物群可以表现出丰富而复杂的行为，并能够做出复杂的集体决定。研究这些相互作用有助于深入了解集体决策、适应性网络和分工。理解这些大规模复杂的相互作用需要新的数据分析方法，这个项目将通过生物学家和计算机科学家之间的跨学科合作来解决这一问题。这项工作的外延部分包括演示和解释生物体群体的复杂行为，以及如何使用计算机研究这些行为。该系统旨在广泛适用，并将通过在线下载该软件的方式传播给其他研究人员。

Across the biological sciences, high-quality, high frame-rate images and video are widely used in analysis of sub-cellular organelles to individual organisms. The use of video has become so pervasive, that many research teams accumulate much more data than can be reasonably analyzed by current methods. This project will develop algorithms to allow the rapid and affordable mining of very large video sets for motion behaviors of interest in the study of the collective behavior and emergent properties of complex systems of organisms. This work is motivated by the needs of domain experts in various areas of biological motion analysis and incorporates research in computer vision, image processing, user interfaces, and visualization. A key part of the project is a fast, semi-automated biological object tracker that leverages the computational power of off-the-shelf hardware and incorporates simple user interactions to dramatically improve tracking accuracy in the types challenging cases that frequently arise in biological image analysis. Domain experts will be able to search motion databases through visual query by selecting an example motion of interest. The system will be evaluated on three biological research applications: honeybee behavior, ant colony networks, and cell motion analysis. Groups of organisms can display a rich and sophisticated behavioral repertoire and are able to make complex collective decisions. Studying these interactions provides insight into collective decision-making, adaptive networks, and division-of-labor. Understanding these complex interactions over large time scales requires new methods of data analysis that this project will address with an interdisciplinary collaboration between biologists and computer scientists. The outreach components of this work include demonstrations and explanations of the complex behavior of groups of organisms and how these behaviors can be studied using computing. The system is intended to be widely-applicable and will be disseminated to other researchers by making the software available for download online.