This set of videos extends the rigid 2D part matching algorithm to 3D, which allows us to introduce techniques for acquiring and processing 3D data, including planar patch extraction and 3D pose estimation. We need to also introduce a simple wire-frame modelling system and adapt the 2D Interpretation Tree matching example. We also introduce 3 more examples of least-square parameter estimation algorithms.
We introduce the idea of a range image and show how we can use it for recognising rigid objects as an extension to the 2D rigid object recognition approach presented previously.
The video covers the main approaches to obtaining a range image, with a focus on triangulation sensors.
We show some example of images acquired using range sensors and point out some of the problems that can arise.
Given a range image, we introduce a region growing approach to extracting the large planes represented in the image. The least square approach used is also suitable for many other linear parameter estimation problems.
By analogy to the 2D wireframe model used to recognise the flat parts, we extend this model to 3D.
By analogy to the 2D flat part matching, we extend the Interpretation Tree to recognise the 3D parts, given the wireframe model and planar patches.
Given a set of model plane : data plane matches, we look at least square methods for estimating the pose that maps the model onto the data, and at methods for verifying that the match and pose estimates are good.
We discuss some of the benefits of range sensing