This article describes a technique that can reliably align arbitrary 2D depictions of an architectural site, including drawings, paintings, and historical photographs, with a 3D model of the site. ...This is a tremendously difficult task, as the appearance and scene structure in the 2D depictions can be very different from the appearance and geometry of the 3D model, for example, due to the specific rendering style, drawing error, age, lighting, or change of seasons. In addition, we face a hard search problem: the number of possible alignments of the painting to a large 3D model, such as a partial reconstruction of a city, is huge. To address these issues, we develop a new compact representation of complex 3D scenes. The 3D model of the scene is represented by a small set of
discriminative visual elements
that are automatically learned from rendered views. Similar to object detection, the set of visual elements, as well as the weights of individual features for each element, are learned in a discriminative fashion. We show that the learned visual elements are reliably matched in 2D depictions of the scene despite large variations in rendering style (e.g., watercolor, sketch, historical photograph) and structural changes (e.g., missing scene parts, large occluders) of the scene. We demonstrate an application of the proposed approach to automatic rephotography to find an approximate viewpoint of historical paintings and photographs with respect to a 3D model of the site. The proposed alignment procedure is validated via a human user study on a new database of paintings and sketches spanning several sites. The results demonstrate that our algorithm produces significantly better alignments than several baseline methods.
This paper introduces an approach to synthetize new CAD assemblies from existing STEP files. The algorithm first generates linkage graph by detecting linkage between components. Then it detects ...linkages similarities between components of different assemblies while analyzing the associated graphs. Finally, it exchanges the similar components. The similarities in a family of components must be formalized by the user. Then the similar components can be replaced by the other through smart placements. This method allows to automatically generate a wide variety of new consistent assemblies sharing the same semantics, in order to create databases of CAD assemblies ready for machine learning applications. It has been validated on several cases.
Product development is an iterative process, partially due to changes in both company internal and external product requirements, resulting in changes to the product under development. These changes ...might require recapitulation of design rationale and result in re-doing assessments and syntheses of different kinds. One way to support this work is to proactively model in such a way that as much as possible of the previous work can be re-used. Not only within one product development project but also across and to future ones. Modelling for re-use can be done by documenting design rationale and formalising performed activities as design guidelines or computer scripts. To be able to find and re-use this information it could be attached to the product features which it relates to. Since geometry is such a core product characteristic, especially within the mechanical industry, and is often modelled as CAD-models, this paper presents a review of CAD-model capabilities and restrictions to serve as a carrier of multidisciplinary information. This is done by; enquiring three Swedish companies, exploring an automated Finite Element Analysis method utilising the CAD-model as a carrier of information, and reviewing different CAD software capabilities. Results show that there are at least seven extension techniques, out of which all are currently being used or considered to be in the future, by at least one company. Further, depending on the extension technique, extendibility and human-comprehension of the added information differ.
The digital manufacturing deviation inspection can be used to assess the dimension, form, and position errors, which is more efficient, more consistent, and more robust than manual inspection. CAD ...models are used as the nominal data and compared with the scanned data of parts to obtain manufacturing errors. CAD model tessellation is a pivotal step in deviation inspection, which directly determines the reliability of analytical results. It requires that the generated triangle mesh has dense vertices, high approximation accuracy, and consistent orientation. A single surface patch tessellation method is proposed, and the adaptive quadtree is first constructed to divide the domain into small spatial grids in the parameter space of the patch; then, a clip-based method is presented to triangulate boundary grids, and inner grids could be triangulated by the constrained Delaunay triangulation method. Furthermore, the mesh patches should be spliced and reoriented consistently, which is very important for deviation analysis. To solve this problem, a robust gap healing method based on merging boundary points of patches is developed, so that all boundary vertices of adjacent patches could completely coincide along the common edges, and all mesh patches would be stitched into a complete mesh without gaps. Finally, the face normal of mesh is adjusted to obtain a consistent orientation. Experiments on industrial CAD models demonstrate that the proposed method could realize faster and more robust mesh generation than the state-of-the-art approaches, including three commercial and two open-source software.
•A geometric model generator for horticultural products is presented.•Geometric models are made, based on 3D non-destructive imaging.•A 3D shape description technique is established.•The generator is ...fast and requires minimal user intervention.•The resulting geometric models show realistic biological variability.
A novel geometric model generator for horticultural products is presented, which generates 3D models of fruits using shape description techniques based on shapes obtained experimentally from a measured dataset of fruits by non-destructive X-ray CT imaging. For this purpose, the 3D contour of each fruit in the scanned dataset was represented with a 2D shape signature, obtained after applying the spherical coordinate transformation. After normalisation, these signatures were described with 2D Fourier descriptors. Statistical analysis of these descriptors for all scanned fruits allowed automatic generation of new geometric fruit models, representative of the measured dataset. The accuracy of the generator was validated by means of the distributions of volumes and surface area to volume ratios of fruit scans and the newly generated shapes. This 3D shape description and generation method allows to process the entire 3D contour of the observed objects and can be applied to all star-shaped objects (shapes that do not curve back on themselves with respect to the centre of mass). This way, more accurate geometrical models can be produced compared to similar model generators based on shape description using 2D cross-sections. This generator enables fast generation of geometrical models to be used in numerical simulations of heat or mass transport phenomena within horticultural products or their exchange processes with the surrounding environment.
CAD assemblies are essentially reduced to a set of components, often solids, and a user-defined tree structure. Given the generation process of CAD assembly models where the user may instantiate ...several times a given component, some solids may occur more than once. If this tree structure incorporates some functional information, this is not mandatory and it cannot be regarded as a reliable functional description. Similarly, component occurrences may not always end up being simple copies of a given solid. To this end, we introduce the concept of intrinsic assembly model and describe and illustrate an associated set of geometry processing operators that can produce an intrinsic shape descriptor of assembly components and extract assembly structure using symmetries, alignments, ... As a complement, it is described how this intrinsic model can become an intrinsic knowledge-based assembly model. Some geometric concepts are mapped to symbolic information using ontology and new structural assembly information is derived using inferences. All these automated processes and mappings enforce the consistency of the proposed model. Illustrative examples show that this model is a first basis toward a functional description of an assembly where new inference rules can be added to express and characterize functional information. A website
http://3dassblyanlysis.gforge.inria.fr/3d/
gives a public access to a knowledge-based assembly example (available with IE and Firefox navigators).
Reusing previous CAD assembly models directly in new product development is almost impossible in One-of-a-Kind Production (OKP) in which customer requirements vary from one to another. As such, ...modularisation of CAD assembly models is required to facilitate modular design for OKP. However, to the authors' best knowledge, there has been no research carried out on modularisation of CAD assembly models so far. To bridge this gap and make the best use of existing CAD models, this paper proposes a novel module partition approach, to group existing CAD assembly models into modules based on component dependencies. In this approach, an extraction algorithm was developed to extract assembly information from a given assembly model directly, by using automated programmable interfaces of CAD software tools. The extracted information is processed to generate the component design structure matrix (DSM) representing hierarchical relations and dependency strengths between components. Four popular hierarchical clustering methods were used to work with the component DSM to produce results of module partition. A case study was carried out to illustrate the proposed methods and demonstrate their feasibility. It enables OKP companies to respond rapidly to changing customer requirements and develop customised products in a short period.
Object detection (OD) is used for visual quality control in factories. Images that compose training datasets are often collected directly from the production line and labeled with bounding boxes ...manually. Such data represent well the inference context but might lack diversity, implying a risk of overfitting. To address this issue, we propose a dataset construction method based on an automated pipeline, which receives a CAD model of an object and returns a set of realistic synthetic labeled images (code publicly available). Our approach can be easily used by non‐expert users and is relevant for industrial applications, where CAD models are widely available. We performed experiments to compare the use of datasets obtained by the two different ways—collecting and labeling real images or applying the proposed automated pipeline—in the classification of five different industrial parts. To ensure that both approaches can be used without deep learning expertise, all training parameters were kept fixed during these experiments. In our results, both methods were successful for some objects but failed for others. However, we have shown that the combined use of real and synthetic images led to better results. This finding has the potential to make industrial OD models more robust to poor data collection and labeling errors, without increasing the difficulty of the training process.
Modern CAD systems generate feature-based product shape models with parameterization and constraints. Until recently, standards for CAD data exchange among different CAD systems were restricted to ...the exchange of pure shape information. These standards ignored the construction history, parameters, constraints, features and other elements of ‘design intent’ present in the model to be transferred. This paper suggests an implementational foundation for CAD data exchange with the preservation of design intent, based on the use of newly published parts of the International Standard ISO 10303 (STEP). Case studies are presented which employ a hypothetical STEP application protocol (AP) using Parts 55, 108 and 111 of ISO 10303. A prototype translator based on this AP has been implemented and tested. The paper reports on the experience gained in ‘intelligent’ data exchange.