Point normal calculation and cloud registration are two of the most common operations in point cloud processing. However, both are vulnerable to issues of numerical precision and loss of significance. This paper documents how loss of significance in the open-source Point Cloud Library can create erroneous point normals and cause cloud registration to fail. Several test clouds are used to demonstrate how the loss of significance is caused by tight point spacing and clouds being shifted far from the origin of their coordinate system. The results show that extreme loss of significance can cause point normals to be calculated with a random orientation, and cause meters of error during cloud registration. Depending on the structure of the point cloud, loss of significance can occur when the cloud is at hundreds or even tens of meters from the origin of its coordinate system. Shifting to larger data types (e.g., from 32-bit “floats” to 64-bit “doubles”) can alleviate the problem but will not solve it completely. Several “best practice” recommendations for avoiding this issue are proposed. But the only solution guaranteed to eliminate loss of significance is de-meaning the entire cloud, or clusters of points before processing.

Visual localization algorithms, i.e., methods that estimate the camera pose of a query image in a known scene, are core components of many applications, including self-driving cars and augmented / mixed reality systems. State-of-the-art visual localization algorithms are structure-based, i.e., they store a 3D model of the scene and use 2D-3D correspondences between the query image and 3D points in the model for camera pose estimation. While such approaches are highly accurate, they are also rather inflexible when it comes to adjusting the underlying 3D model after changes in the scene. Structureless localization approaches represent the scene as a database of images with known poses and thus offer a much more flexible representation that can be easily updated by adding or removing images. Although there is a large amount of literature on structure-based approaches, there is significantly less work on structureless methods. Hence, this paper is dedicated to providing the, to the best of our knowledge, first comprehensive discussion and comparison of structureless methods. Extensive experiments show that approaches that use a higher degree of classical geometric reasoning generally achieve higher pose accuracy. In particular, approaches based on classical absolute or semi-generalized relative pose estimation outperform very recent methods based on pose regression by a wide margin. Compared with state-of-the-art structure-based approaches, the flexibility of structureless methods comes at the cost of (slightly) lower pose accuracy, indicating an interesting direction for future work.

Publication in the conference proceedings of EUSIPCO, Lisbon, Portugal, 2014