Philippe Véron

Semantic 3D Media and Content: A semantic-based platform for the digital analysis of architectural heritage

This essay focuses on the fields of architectural documentation and digital representation. We present a research paper concerning the development of an information system at the scale of architecture, taking into account the relationships that can be established between the representation of buildings (shape, dimension, state of conservation, hypothetical restitution) and heterogeneous information about various fields (such as the technical, the documentary or still the historical one). The proposed approach aims to organize multiple representations (and associated information) around a semantic description model with the goal of defining a system for the multi-field analysis of buildings.

Filling holes in meshes using a mechanical model to simulate the curvature variation minimization

The presence of holes in a triangle mesh is classically ascribed to the deficiencies of the point cloud acquired from a physical object to be reverse engineered. This lack of information results from both the scanning process and the object complexity. The consequences are simply not acceptable in many application domains (e.g. visualization, finite element analysis or STL prototyping). This paper addresses the way these holes can be filled in while minimizing the curvature variation between the surrounding and inserted meshes. The curvature variation is simulated by the variation between external forces applied to the nodes of a linear mechanical model coupled to the meshes. The functional to be minimized is quadratic and a set of geometric constraints can be added to further shape the inserted mesh. In addition, a complete cleaning toolbox is proposed to remove degenerated and badly oriented triangles resulting from the scanning process.

Geometric and form feature recognition tools applied to a design for assembly methodology

The paper presents geometric tools for an automated Design for Assembly (DFA) assessment system. For each component in an assembly a two step features search is performed: firstly (using the minimal bounding box) mass, dimensions and symmetries are identified allowing the part to be classified, according to DFA convention, as either rotational or prismatic; secondly form features are extracted allowing an effective method of mechanised orientation to be determined. Together these algorithms support the fuzzy decision support system, of an assembly-orientated CAD system known as FuzzyDFA.

An Iconography-Based Modeling Approach for the Spatio-Temporal Analysis of Architectural Heritage

The study of historic buildings is usually based on the collection and analysis of iconographic sources such as photographs, drawings, engravings, paintings or sketches. This paper describes a methodological approach to make use of the existing iconographic corpus for the analysis and the 3D management of building transformations. Iconography is used for different goals. Firstly, its a source of geometric information (image-based-modeling of anterior states); secondly, its used for the re-creation of visual appearance (image-based texture extraction); thirdly its a proof of the temporal distribution of shape transformations(spatio-temporal modeling); finally it becomes a visual support for the study of building transformations (visual comparison between different temporal states). The aim is to establish a relation between the iconography used for the hypothetical reconstruction and the 3D representation that depends on it. This approach relates to the idea of using 3D representations like visualization systems capable of reflecting the amount of knowledge developed by the study of a historic building.

Towards recovery of complex shapes in meshes using digital images for reverse engineering applications

When an object owns complex shapes, or when its outer surfaces are simply inaccessible, some of its parts may not be captured during its reverse engineering. These deficiencies in the point cloud result in a set of holes in the reconstructed mesh. This paper deals with the use of information extracted from digital images to recover missing areas of a physical object. The proposed algorithm fills in these holes by solving an optimization problem that combines two kinds of information: (1) the geometric information available on the surrounding of the holes, (2) the information contained in an image of the real object. The constraints come from the image irradiance equation, a first-order non-linear partial differential equation that links the position of the mesh vertices to the light intensity of the image pixels. The blending conditions are satisfied by using an objective function based on a mechanical model of bar network that simulates the curvature evolution over the mesh. The inherent shortcomings both to the current hole-filling algorithms and the resolution of the image irradiance equations are overcome.

Repairing triangle meshes built from scanned point cloud

The Reverse Engineering process consists of a succession of operations that aim at creating a digital representation of a physical model. The reconstructed geometric model is often a triangle mesh built from a point cloud acquired with a scanner. Depending on both the object complexity and the scanning process, some areas of the object outer surface may never be accessible, thus inducing some deficiencies in the point cloud and, as a consequence, some holes in the resulting mesh. This is simply not acceptable in an integrated design process where the geometric models are often shared between the various applications (e.g. design, simulation, manufacturing). In this paper, we propose a complete toolbox to fill in these undesirable holes. The hole contour is first cleaned to remove badly-shaped triangles that are due to the scanner noise. A topological grid is then inserted and deformed to satisfy blending conditions with the surrounding mesh. In our approach, the shape of the inserted mesh results from the minimization of a quadratic function based on a linear mechanical model that is used to approximate the curvature variation between the inner and surrounding meshes. Additional geometric constraints can also be specified to further shape the inserted mesh. The proposed approach is illustrated with some examples coming from our prototype software.

On the use of Machine Learning to Defeature CAD Models for Simulation

Numerical simulations play more and more important role in product development cycles and are increasingly complex, realistic and varied. CAD models must be adapted to each simulation case to ensure the quality and reliability of the results. The defeaturing is one of the key steps for preparing digital model to a simulation. It requires a great skill and a deep expertise to foresee which features have to be preserved and which features can be simplified. This expertise is often not well developed and strongly depends of the simulation context. In this paper, we propose an approach that uses machine learning techniques to identify rules driving the defeaturing step. The expertise knowledge is supposed to be embedded in a set of configurations that form the basis to develop the processes and find the rules. For this, we propose a method to define the appropriate data models used as inputs and outputs of the learning techniques.

Lattice structure lightweight triangulation for additive manufacturing

Abstract Additive manufacturing offers new available categories of geometries to be built. Among those categories, one can find the well developing field of lattice structures. Attention has been paid on lattice structures for their lightweight and mechanical efficiency ratio, thus leading to more optimized mechanical parts for systems. However this lightness only holds true from a mass related point of view. The files sent to additive manufacturing machines are quite large and can go up to such sizes that machines can freeze and get into malfunction. This is directly related to the lattice structures tendency to be of a high geometric complexity. A large number of vertices and triangles are necessary to describe them geometrically, thus leading to larger file sizes. With the increasing use of lattice structures, the need for their files to be lighter is also rising. This paper aims at proposing a method for tessellating a certain category of such structures, using topologic and geometric criteria to generate as few as possible triangles, thus leading to lightweight files. The triangulation technique is driven by a chordal error that controls the deviation between the exact and tessellated structures. It uses interpolation, boolean as well as triangulation operators. The method is illustrated and discussed through examples from our prototype software.

Towards an Ontology for Annotating Degradation Phenomena

In the field of built heritage, a huge amount of data describes the state of monuments: documentary data (texts, sounds, images) as well as analytic data from sensors, provides historical, archeological and constructive information. These data, produced by experts coming from several fields, are the foundation for the creation of new information through scientific observations. The problem is that these data are not related nor spatialized. Nowadays we are able to generate extreme accurate 3D model with throughout images-based or laser scanner acquisition. However these 3D models do not carry information regarding their morphological complexities. The design of an ontology for the conservation domain seems to be the best solution in order to obtain understandable entities thanks to their own data related between them. This article present the first attempt of a development of a web information system based on spatialized images semantic annotations tool, related to a domain ontology describing knowledge regarding stone degradation phenomena. Our approach is to produce a domain ontology able to document and therefore provide a framework able to help the decision-making process of experts in the cultural heritage conservation domain. The information annotated with the use of this ontology can enrich not only the scientific observations, but also to help to create new knowledge. In this way, it is possible to link and gather quantitative and qualitative aspect into only one information system. We will propose several example and queries able to exploit the reasoning power of the above information system.

An approach for precise 2D/3D semantic annotation of spatially-oriented images for in situ visualization applications

Thanks to nowadays technologies, innovative tools afford to increase our knowledge of historic monuments, in the field of preservation and valuation of cultural heritage. These tools are aimed to help experts to create, enrich and share information on historical buildings. Among the various documentary sources, photographs contain a high level of details about shapes and colors. With the development of image analysis and image-based-modeling techniques, large sets of images can be spatially oriented towards a digital mock-up. For these reasons, digital photographs prove to be an easy to use, affordable and flexible support, for heritage documentation. This article presents, in a first step, an approach for 2D/3D semantic annotations in a set of spatially-oriented photographs (whose positions and orientations in space are automatically estimated). In a second step, we will focus on a method for displaying those annotations on new images acquired by mobile devices in situ. Firstly, an automated image-based reconstruction method produces 3D information (specifically 3D coordinates) by processing a large images set. Then, images are semantically annotated and a process uses the previously generated 3D information inherent to images for the annotations transfer. As a consequence, this protocol provides a simple way to finely annotate a large quantity of images at once instead of one by one. As those images annotations are directly inherent to 3D information, they can be stored as 3D files. To bring up on screen the information related to a building, the user takes a picture in situ. An image processing method allows estimating the orientation parameters of this new photograph inside the already oriented large images base. Then the annotations can be precisely projected on the oriented picture and send back to the user. In this way a continuity of information could be established from the initial acquisition to the in situ visualization.