3D object recognition and tracking from live cameras

这一合作研究项目的主要目标是确定和验证用户与软件系统交互的新模式。更具体地说，使用对象识别组件，用户将能够通过使用识别的3D对象与软件产品交互。目标是能够从给定的3D对象中提取出具有代表性的紧凑模型。该模型将由不同的特征组成，然后必须将这些特征与从传入视频帧中提取的特征相匹配。一旦建立了匹配，就必须推断对象的姿势。这里的挑战是能够以帧速率执行健壮的模型匹配，从而创建无缝的交互体验。模型构建步骤也必须简单，以便非专家用户可以轻松地将新对象添加到系统中。

The main objective of this collaborative research project is to identify and validate new modalities for users to interact with software systems. More specifically, using an object recognition component, users will be able to interact with a software product by using a recognized 3D object. The objective is to be able to extract a representative and compact model from a given 3D object. This model will be composed of distinct features that must then be matched to the ones extracted from an incoming video frames. Once a match has been established, the object pose then has to be inferred. The challenge here is to be able to perform robust model matching at frame rate such to create a seamless interactive experience. The model building step must also be simple such that non-expert users can add new objects to the system in an easy way.