Guy Godin

Detailed 3D reconstruction of large-scale heritage sites with integrated techniques

Many cultural heritage applications require 3D reconstruction of real-world objects and scenes. Over the past few years, it has become increasingly common to use 3D digitization and modeling for this purpose. This is mainly due to advances in laser-scanning techniques, 3D modeling software, image-based modeling techniques, computer power, and virtual reality. Our approach integrates several technologies based on our experience over more than a decade of trying to accurately and completely model large-scale heritage monuments and sites. Using both interactive and automatic techniques, we can model a highly detailed structure or site at various levels of detail. We use image-based modeling for basic shape and structural elements, and laser scanning for fine details and sculpted surfaces. To present the site in its proper context, we use image-based rendering for landscapes and surroundings. To apply this approach, we created hundreds of models from sites all over the world for documentation, walk-through movies, and interactive visualization. The results were compelling and encouraging.

Separation of diffuse and specular components of surface reflection by use of polarization and statistical analysis of images

The image of an opaque object is created by observing the reflection of the light incident on its surface. The dichromatic reflection model describes the surface reflection as the sum of two components, diffuse and specular terms. The specular reflection component is usually strong in its intensity and polarized significantly compared to the diffuse components. On the other hand, the intensity of the diffuse component is weak and it tends to be unpolarized except near occluding contours. Thus, the observation of an object through a rotating polarizer approximately yields images containing constant diffuse component and specular component of different intensity. In this paper, we show that diffuse and specular components of surface reflection can be separated as two independent components when we apply independent component analysis to the images observed through a polarizer of different orientations. We give a separation simulation of artificial data and also give some separation results of real scenes.

Color reflectance modeling using a polychromatic laser range sensor

A system for simultaneously measuring the 3-D shape and color properties of objects is described. Range data are obtained by triangulation over large volumes of the scene, whereas color components are separated by means of a white laser. Details are given concerning the modeling and the calibration of the system for bidirectional reflectance-distribution functions measurements. A reflection model is used to interpret the data collected with the system in terms of the underlying physical properties of the target. These properties are the diffuse reflectance of the body material, the Fresnel reflectance of the air media interface, and the slope surface roughness of the interface. Experimental results are presented for the extraction of these parameters. By allowing the subtraction of highlights from color images and the compensation for surface orientation, spectral reflectance modeling can help to understand 3-D scenes. A practical example is given where a color and range image is processed to yield uniform regions according to material pigmentation. >

A texture-mapping approach for the compression of colored 3D triangulations

We present an algorithm that constructs compact and realistic descriptions of colored 3D objects using texture mapping on compressed triangulations. A high-resolution triangular mesh model is created by integrating measurements from a color 3D laser sensor. Each vertex is attributed with a RGB color value. The high-resolution triangulation is transformed into a compressed triangulation and a texture map. This map embeds the color information of the vertices removed during the geometric compression and projected on the lower resolution triangulation. We describe the algorithm for the rapid and efficient construction of a texture map for compressed triangulations of arbitrary topology. Experiments show that high compression rates can be achieved while maintaining good visual similarity between the original and compressed models.

Object model creation from multiple range images: acquisition, calibration, model building and verification

This paper demonstrates the accuracy of a prototype Laser Range Camera (LRC) developed at the National Research Council of Canada for the creation of models of real objects. A laser survey performed in collaboration with the Canadian Space Agency and NASA is used as a test case. The object selected for this particular test case is the Orbiter Docking System (ODS) located at the Kennedy Space Center, Florida. During the laser survey, 128 range (and registered intensity) images were acquired all around the ODS. These images were then processed in our laboratory. A full model of the top portion of the ODS was created along with an almost complete model of the ODS. The ODS has a diameter of 1.6 m and a height of 3.9 m. Targets mounted on the top portion of the ODS were used to assess the accuracy of the calibration and of the image registration process. These targets were measured with a network of theodolites a day prior to the laser survey and used as a reference. With the current calibration and range image registration techniques, an accuracy better than 0.25 mm in X and Y, and, 0.80 mm in Z was achieved. These results compare favorably with the single point accuracy obtained after calibration, i.e., about 0.25 mm in X and Y, and, 0.50 mm in Z. These figures and others should testify on the usefulness of a LRC for accurate model building.

A nearest neighbor method for efficient ICP

A novel solution is presented to the Nearest Neighbor Problem that is specifically tailored for determining correspondences within the Iterative Closest Point Algorithm. The reference point set P is preprocessed by calculating for each point p/spl I.oarr//sub i//spl isin/P that neighborhood of points which lie within a certain distance /spl epsiv/ of p/spl I.oarr//sub i/. The points within each /spl epsiv/-neighborhood are sorted by increasing distance to their respective p/spl I.oarr//sub i/. At runtime, the correspondences are tracked across iterations, and the previous correspondence is used as an estimate of the current correspondence. If the estimate satifies a constraint, called the Spherical Constraint, then the nearest neighbor falls within the /spl epsiv/-neighborhood of the estimate. A novel theorem, the Ordering Theorem, is presented which allows the Triangle Inequality to efficiently prune points from the sorted /spl epsiv/-neighborhood from further consideration. The method has been implemented and is demonstrated to be more efficient than both the k-d tree and Elias methods. After /spl sim/40 iterations, fewer than 2 distance calculations were required on average per correspondence, which is close to the theoretical minimum of 1. Furthermore, after 20 iterations the time expense per iteration was demonstrated to be negligibly more than simply looping through the points.

Active optical 3D imaging for heritage applications

High-resolution 3D imaging and modeling is an important application in the heritage field. We describe several demonstration projects conducted in collaboration with museums and conservation agencies.

A method for the registration of attributed range images

Registration of range images requires the identification of common portions of surfaces between which a distance minimization is performed. This paper proposes a framework for the use of dense attributes of range image elements as a matching constraint in the registration. These attributes are chosen to be invariant to rigid transformations, so that their value is similar in different views of the same surface portion. Attributes can be derived from the geometry information in the range image, such as surface curvature, or be obtained from associated intensity measurements. The method is based on the Iterative Closest Compatible Point algorithm augmented with a random sampling scheme that uses the distribution of attributes as a guide for point selection. Distance minimization is performed only between pairs of points considered compatible on the basis of their attributes. The performance of the method is illustrated on a rotationally symmetric object with color patterns.

GoLD: interactive display of huge colored and textured models

This paper presents a new technique for fast, view-dependent, real-time visualization of large multiresolution geometric models with color or texture information. This method uses geomorphing to smoothly interpolate between geometric patches composing a hierarchical level-of-detail structure, and to maintain seamless continuity between neighboring patches of the model. It combines the advantages of view-dependent rendering with numerous additional features: the high performance rendering associated with static preoptimized geometry, the capability to display at both low and high resolution with minimal artefacts, and a low CPU usage since all the geomorphing is done on the GPU. Furthermore, the hierarchical subdivision of the model into a tree structure can be accomplished according to any spatial or topological criteria. This property is particularly useful in dealing with models with high resolution textures derived from digital photographs. Results are presented for both highly tesselated models (372 million triangles), and for models which also contain large quantities of texture (200 million triangles + 20 GB of compressed texture). The method also incorporates asynchronous out-of-core model management. Performances obtained on commodity hardware are in the range of 50 million geomorphed triangles/second for a benchmark model such as Stanfords St. Matthew dataset.

Sensors and algorithms for the construction of digital 3-D colour models of real objects

This paper describes sensors and algorithms developed and used for the creation of coloured 3-D triangular meshes from a set of range images with registered colour measurements. The objective is the creation of a digital model of a real object that is compatible with established computer graphics techniques. Triangular meshes are the fundamental rendering primitive supported by most high performance graphics workstations. The surface colour of objects can be represented by attributing reflectance values to the vertices of the mesh. Such a geometric and reflectance model provides a uniform and general representation for sculptured surfaces with non-uniform colouring.