The most recent revision of this document will always be found at

The paper introduces a data collection system and a processing pipeline for automatic geo-registered 3D reconstruction of urban scenes from video. The system collects multiple video streams, as well as GPS and INS measurements in order to place the reconstructed models in georegistered coordinates. Besides high quality in terms of both geometry and appearance, we aim at real-time performance. Even though our processing pipeline is currently far from being real-time, we select techniques and we design processing modules that can achieve fast performance on multiple CPUs and GPUs aiming at real-time performance in the near future. We present the main considerations in designing the system and the steps of the processing pipeline. We show results on real video sequences captured by our system. 1

This paper describes a recently developed toolkit for creating environmental maps of indoor environments from image sequences. First, the chosen map representation – texture-mapped planes located in a single coordinate frame – is placed within the broader context of appearance-based and sparse feature-based visual mapping approaches. Drawing from the well-developed field of structure-from-motion, an algorithm is presented for building such maps with a limited sensor suite: monocular vision and odometry. Maps validating this approach are presented, along with task-independent metrics made possible by the representation. We evaluate our maps according to these metrics and conclude with implications for future sensor-limited spatial reasoning.

Digitally documenting complex heritage sites such as castles is a desirable yet difficult task with no established framework.Although 3D digitizing and modelling with laser scanners, Photogrammetry, and computer aided architectural design (CAAD) are maturing, each alone is inadequate to model an entire castle in details.We present a sequential approach that combines multiple techniques, each where best suited, to capture and model the fine geometric detail of castles.We provide new contributions in several areas: an effective workflow for castle 3D modelling, increasing the level of automation and the seamless integration of models created independently from different data sets.We tested the approach on various castles in Northern Italy and the results demonstrated that it is effective, accurate, and creates highly detailed models suitable for interactive visualization. It is also equally applicable to other types of large complex architectures. � Figure 1: Examples of castle structures and locations.

The paper introduces a data collection system and a processing pipeline for automatic geo-registered 3D reconstruction of urban scenes from video. The system collects multiple video streams, as well as GPS and INS measurements in order to place the reconstructed models in georegistered coordinates. Besides high quality in terms of both geometry and appearance, we aim at real-time performance. Even though our processing pipeline is currently far from being real-time, we select techniques and we design processing modules that can achieve fast performance on multiple CPUs and GPUs aiming at real-time performance in the near future. We present the main considerations in designing the system and the steps of the processing pipeline. We show results on real video sequences captured by our system.

We present a novel approach on digitizing large scale unstructured environments like archaeological excavations using off-the-shelf digital still cameras. The cameras are calibrated with respect to few markers captured by a theodolite system. Having all cameras registered in the same coordinate system enables a volumetric approach. Our new algorithm has as input multiple calibrated images and outputs an occupancy voxel space where occupied pixels have a local orientation and a confidence value. Both, orientation and confidence facilitate an efficient rendering and texture mapping of the resulting point cloud. Our algorithm combines the following new features: Images are backprojected to hypothesized local patches in the world and correlated on these patches yielding the best orientation. Adjacent cameras build tuples which yield a product of pairwise correlations, called strength. Multiple camera tuples compete eachother for the best strength for each voxel. A voxel is regarded as occupied if strength is maximum along the normal. Unlike other multi-camera algorithms using silhouettes, photoconsistency, or global correspondence, our algorithm makes no assumption on camera locations being outside the convex hull of the scene. We present compelling results of outdoors excavation areas. 1.