Advanced search
Start date
Betweenand


Unorganized point cloud segmentation using watershed transform

Full text
Author(s):
Pedro Victor Vieira de Paiva
Total Authors: 1
Document type: Master's Dissertation
Press: Limeira, SP.
Institution: Universidade Estadual de Campinas (UNICAMP). Faculdade de Tecnologia
Defense date:
Examining board members:
Eloisa Dezen Kempter; Hélio Pedrini; André Leon Sampaio Gradvohl
Advisor: Marco Antonio Garcia de Carvalho; Eloisa Dezen Kempter
Abstract

The main challenge for historical constructions conservation is to understand its real state. Digital representation demand leads to the application of new acquisition technologies. Geometric and texture data obtained through multi-sensors presents as promising solutions. Such sensors, like 3D laser scanners and photogrammetric methods, commonly generate point cloud structures. However, the large volume of data generated by those techniques makes further analysis arduous. This work extends the ability of the Watershed Transform in the segmentation of point clouds representing buildings in their architectural components. In addition to the change of the object in which the transform is usually applied (images), the method is enriched with geometric information by means of curvature analysis, making the approach hybrid. This dissertation also proposes a new strategy for the application of morphological operators in clouds of unorganized points by means of the adjacency of voxels in octree, besides the definition of a protocol of acquisition of clouds of construction points. Finally, it was created from a dataset of Brazilian historical patrimonies labeled in architectural elements by specialists. We present a set of experiments that evaluate the method proposed by Precision-Recall metrics, in addition to comparisons with techniques consolidated in the state of the art (AU)

FAPESP's process: 17/02787-9 - A framework to integrate multi-sensors data with building information modeling to support historic assets conservation and management
Grantee:Pedro Victor Vieira de Paiva
Support Opportunities: Scholarships in Brazil - Master