IMAP: Image-based modeling with autonomous positioning

点击翻译按钮获取中文摘要

Over the last 20 years, optical 3D object digitizing systems have evolved significantly. It is currently possible to digitize the shape of a wide range of objects with sub-millimetre accuracy and resolution. This type of vision-based technology has significantly contributed to various fields such as industrial inspection, prototyping, design and arts, tailor-made ortheses, archaeology documentation and, more generally, to simulation. Canada is among the world leaders in this field and several service providers and equipment manufacturers have blossomed in the last 10 years. Currently, the predominant challenges faced by designers consist in improving the flexibility of the digitizing systems and achieving better quality in capturing surface appearance. These challenges raise fundamental problems in the field of computer vision. First, developing an intelligent auto-referencing system based on image content should deal concurrently with capturing and integrating hundreds, and even thousands of images that must be referenced to each other. The system should provide a diagnosis of the positioning quality based on natural and artificial target fiducials. This self-positioning system will then be exploited for facilitating geometry recovery in addition to serving as a tool for surface appearance capture. Instead of capturing the texture from a limited set of images, it is expected to make interactive capture of light fields and reflectance properties of objects possible for obtaining incomparable quality with respect to current texture capture and mapping methods. This project will contribute to the training of 4 students at the doctorate level and a few undergraduate students. The project is supported by a company specialized in 3D digitizing as well as a research team in mobile robotics at Defence Research and Development Canada.