Carla Nardinocchi

Complete classification of raw LIDAR data and 3D reconstruction of buildings

LIDAR (LIght Detection And Ranging) data are a primary data source for digital terrain model (DTM) generation and 3D city models. This paper presents a three-stage framework for a robust automatic classification of raw LIDAR data as buildings, ground and vegetation, followed by a reconstruction of 3D models of the buildings. In the first stage the raw data are filtered and interpolated over a grid. In the second stage, first a double raw data segmentation is performed and then geometric and topological relationships among regions resulting from segmentation are computed and stored in a knowledge base. In the third stage, a rule-based scheme is applied for the classification of the regions. Finally, polyhedral building models are reconstructed by analysing the topology of building outlines, building roof slopes and eaves lines. Results obtained on data sets with different ground point density, gathered over the town of Pavia (Italy) with Toposys and Optech airborne laser scanning systems, are shown to illustrate the effectiveness of the proposed approach.

Building extraction from LIDAR data

A strategy for building reconstruction relying on LIDAR data only is presented. Roofs are modeled as plane surfaces, connected along ridges and bordered by the eaves lines. Edge pixels and plane surfaces are detected and labelled as roof slopes based on gradient orientation and plane fitting by RANSAC; a similar procedure applies to eaves lines. Based on geometric reasoning, the topology of the roof slopes and walls is reconstructed, deriving also roof ridges and roof corners. Constraints to be enforced in order to obtain a building shape consistent with the geometric evidence are established. The final reconstruction is performed by a global l.s. adjustment where all raw data contribute to a determination of the building faces while the constraints are satisfied. Preliminary results from a laser scanning survey with a ground resolution of 1 m are presented.

Where is photogrammetry heading to? State of the art and trends

The objective of this paper is to highlight current trends in photogrammetry, trying to foresee where they will lead the discipline in the next years. To this aim, first some remarks on the challenges brought to photogrammetry by other sensors and a brief historical survey of some research topics, where an increasing convergence between photogrammetry and computer vision is apparent, will be presented. Then, a necessarily concise review of the advances in automation in three basic photogrammetric tasks (namely image orientation, surface reconstruction and object restitution) will be illustrated. The purpose of the review is to highlight how the fruitful dialog between photogrammetry and computer vision led to today’s achievements and to point out what kind of approaches seem to be winning in the search for viable and robust solutions in the automation of processes. Finally, the conclusions will look at this convergence in the perspective of academic career.

Automatic extraction of LIDAR data classification rules

LIDAR (Light Detection And Ranging) data are a primary data source for digital terrain model (DTM) generation and 3D city models. This paper presents an AdaBoost algorithm for the identification of rules for the classification of raw LIDAR data mainly as buildings, ground and vegetation. First raw data are filtered, interpolated over a grid and segmented. Then geometric and topological relationships among regions resulting from segmentation constitute the input to the tree-structured classification algorithm. Results obtained on data sets gathered over the town of Pavia (Italy) are compared with those obtained by a rule-based approach previously presented by the authors for the classification of the regions.

Web Spherical Panorama for Cultural Heritage 3D Modeling

Among the several information available on web the spherical panoramas by Google can be an important reference for the 3D models elaboration of urban contexts and historic buildings, difficult to access or even destroyed as a result of wars or natural disasters. Starting from the spherical photogrammetry, thoroughly tested and used by some scholars, the goal of this research was to verify the reliability in the use of spherical panoramas by Google in the architectural and urban survey, considered that in some cases such as in the recent earthquake that hit central Italy, these images represent the only “visual” evidence of something that no longer exists.

Design of a Passive System for Human Body Reconstruction in the Fashion Industry

The fashion industry is looking with interest at the progress in 3D human body modelling, to streamline and optimize the production of customized, on-demand clothes. The accuracy required for length measurements of a 3D body model is in the order of magnitude of a few millimetres. The paper presents the first results of a study for the development of a body scanning system for the apparel and the fashion industry based on photogrammetric techniques. With the objective to anticipate at least some of the practical aspects of the study, an investigation into the accuracy and completeness of the body model reconstruction as a function of camera network configuration, type of image matching algorithm and camera orientation has been performed.