Ferrograph analysis has been adopted over decades for determining the root causes of on-going wear faults. After decades of manual operation, this traditional technique is being driven by intelligent algorithms for automatic identification of wear debris. However, the accuracy and robustness of this algorithm remain marginalized when applied in industry due to various types and color blurry of particles. To address this issue, this paper introduces an automatic ferrograph analysis model with a segmentation method and a two-level classification strategy. In order to obtain wear particles from the color ferrograph image, an adaptive Otsu threshold is adopted in three channel images to solve the color blurry in particle segmentation. By grouping particle parameters into shape and morphology ones, a two-level identification strategy is proposed. The first one is to classify rubbing, cutting and spherical particles, referring to the fuzzy approach degree of shape parameters. In the second level, the shape-close particles are classified with imperceptible textures and back propagation Neural Network (BPNN). These objective parameters are constructed by applying the principal component analysis into seven texture features, and inputted into a BPNN-based model to classify fatigue and severe sliding particles. In order to train the BPNN, more than 100 ferrograph images are sampled together, whereby standard ferrograph analysis is performed on the particle identification. The performance of the identification exhibits an accuracy exceeding 90% for rubbing, cutting and spherical particles, whereas about 80% accuracy has been registered for both severe sliding and fatigue particles.

The paper describes the design of a wearable and wireless system that allows the real-time identification of some gestures performed by basketball players. This system is specifically designed as a support for coaches to track the activity of two or more players simultaneously. Each wearable device is composed of two separate units, positioned on the wrists of the user, connected to a personal computer (PC) via Bluetooth. Each unit comprises a triaxial accelerometer and gyroscope, a microcontroller, installed on a TinyDuino platform, and a battery. The concept of activity recognition chain is investigated and used as a reference for the gesture recognition process. A sliding window allows the system to extract relevant features from the incoming data streams: mean values, standard deviations, maximum values, minimum values, energy, and correlations between homologous axes are calculated to identify and differentiate the performed actions. Machine learning algorithms are implemented to handle the recognition phase.

This paper presents a new method for extracting feature edges from computer-aided design (CAD)-generated triangulations. The major advantage of this method is that it tends to extract feature edges along the centroids of the fillets rather than along the edges where fillets are connected to nonfillet surfaces. Typical industrial models include very small-radius fillets between relatively large surfaces. While some of those fillets are necessary for certain types of analyses, many of them are irrelevant for many other types of applications. Narrow fillets are unnecessary details for those applications and cause numerous problems in the downstream processes. One solution to the small-radius fillet problem is to divide the fillets along the centroid and then merge each fragment of the fillet with nonfillet surfaces. The proposed method can find such fillet centroids and can substantially reduce the adverse effects of such small-radius fillets. The method takes a triangulated geometry as input and first simplifies the model so that small-radius, or “small,” fillets are collapsed into line segments. The simplification is based on the normal errors and therefore is scale-independent. It is particularly effective for a shape that is a mix of small and large features. Then, the method creates segmentation in the simplified geometry, which is then transformed back to the original shape while maintaining the segmentation information. The groups of triangles are expanded by applying a region-growing technique to cover all triangles. The feature edges are finally extracted along the boundaries between the groups of triangles.

This paper is arranged in three main sections: the first section is a hierarchical method based on clustering and a fuzzy membership system where the tessellated three-dimensional (3D) models are classified into their containing primitives: cylinder, cone, sphere, and flat. In the second section, automated assembly planning (AAP) is considered as the main application of our novel hierarchical primitive classification approach. The classified primitives obtained from the first section are used to define the removal directions between mating parts in an assembly model. Finally, a fuzzification method is used to express the uncertainty of the detected connections between every pair of parts. The acquired uncertainties are used in a user interaction process to approve, deny, or modify the connections with higher uncertainties.

Freehand sketching is an integral part of early design process. Recent years have seen an increased interest in supporting sketching in computer-based design systems. In this paper, we present finite element analysis made easy (FEAsy), a naturalistic environment for static finite element analysis. This tool allows users to transform, simulate, and analyze their finite element models quickly and easily through freehand sketching. A major challenge here is to beautify freehand sketches, and to this extent, we present a domain-independent, multistroke, multiprimitive method which automatically detects and uses the spatial relationships implied in the sketches for beautification. Further, we have also developed a domain-specific rules-based algorithm for recognizing commonly used symbols in finite element analysis (FEA) and a method for identifying different contexts in finite element modeling through combined interpretation of text and geometry. The results of the user study suggest that our proposed algorithms are efficient and robust. Pilot users found the interface to be effective and easy to use.

Free-form surface features (FFSFs) extraction is one of the key issues for redesigning and reediting the surface models exported from commercial software or reconstructed by reverse engineering. In this paper, a coarse-to-fine method is proposed to robustly extract the FFSFs. First, by iterative Laplacian smoothing, a set of height functions are generated, and principal component analysis (PCA) is employed to obtain the appropriate iteration number for the feature field extraction that is then accomplished by the Gaussian mix model (GMM) with a high segmentation threshold. Second, based on the feature field, an adaptive smooth ratio for each vertex is proposed for Laplacian smoothing, which is implemented to generate a precise base surface. Thereby, with the base surface, the FFSFs can be easily extracted by using the GMM. The empirical results illustrate that the proposed method yields improved performance for extracting FFSFs compared with conventional methods.

The problem of segmenting the object from the background is addressed in the proposed Gaussian and Gabor Filter Approach (GGFA) for object segmentation. An improved and efficient approach based on Gaussian and Gabor Filter reads the given input image and performs filtering and smoothing operation. The region occupied by the object is extracted from the image by performing various operations like bilateral filtering, Edge detection, Clustering, and Region growing. The proposed approach experimented on standard images taken from Caltech datasets, Corel Photo CDs, and Weizmann horse datasets show significantly improved results.

Mesh generation from X-ray computed tomography (CT) images of mechanical parts is an important consideration in industrial application, and boundary surface meshes in multimaterial parts can be extracted by generating segmented meshes from segmented images. In this paper, the authors outline a new approach for achieving segmented mesh generation. The image is first subjected to centroidal Voronoi tessellation and Delaunay tessellation steered by a density map to create a triangular mesh while maintaining discontinuities between materials. Given an input domain and a number of initial sites, the energy function is minimized automatically by iteratively updating the Voronoi tessellation and relocating sites to produce optimized domain discretization and form the mesh. Thus, the mesh in question is effectively and quickly segmented into different parts via this new graph cut method. The proposed approach is considered more efficient because there are fewer triangles than pixels, which reduces computation time and memory usage.

We present a robust and efficient approach to directly slicing implicit solids. Different from prior slicing techniques that reconstruct contours on the slicing plane by tracing the topology of intersected line segments, which is actually not robust, we generate contours by a topology guaranteed contour extraction on binary images sampled from given solids and a subsequent contour simplification algorithm which has the topology preserved and the geometric error controlled. The resultant contours are free of self-intersection, topologically faithful to the given r-regular solids and with shape error bounded. Therefore, correct objects can be fabricated from them by rapid prototyping. Moreover, since we do not need to generate the tessellated B-rep of given solids, the memory cost our approach is low—only the binary image and the finest contours on one particular slicing plane need to be stored in-core. Our method is general and can be applied to any implicit representations of solids.

In this paper, we propose an effective solution to reconstruct solid models of existing objects. Specifically, we convert the model reconstruction problem into the issue of feature parameter extraction, and thereby design diverse methods to extract the parameters of basic design features from input surface meshes. After extracting the feature parameters, the corresponding features are constructed. By performing modeling operations on those features, the final solid model is constructed, and meanwhile the complete history of the model building operations is recorded. By introducing the concepts of “feature,” “constraint,” and “modeling history” into the reconstruction process, the design intent is captured and hence represented in the reconstructed model. As a result, the model is geometrically accurate and topologically consistent, and moreover it is flexibly editable, which makes it convenient to carry out model redesign and modification for the innovation applications. A variety of experimental results demonstrate the effectiveness and robustness of this solution.

In this work, we introduce a method named creeping contours for image segmentation into component parts for the purpose of extracting the boundary surfaces of these parts. Creeping contours are contours that expand following a speed function defined by the gradient and curvature at contour points, starting from an initial contour position defined either manually or automatically. Contours in the image creep simultaneously at different speeds, while labels are assigned to contour pixels by the defined creeping condition. We also demonstrate the effectiveness of the proposed method by segmenting 2D grayscale images and 3D volumetric computed tomography images of mechanical parts into multiple segments and generating the boundary surfaces of these parts.

This paper describes a method of machining feature recognition from a freeform surface based on the relationship between unique machining patches and critical points on a component’s surface. The method uses Morse theory to extract critical surface points by defining a scalar function on the freeform surface. Features are defined by region growing between the critical points using a tool path generation algorithm. Several examples demonstrate the efficiency of this approach. The recognized machining features can be directly utilized in a variety of downstream computer aided design/computer aided manufacturing (CAM) applications, such as the automated machining process planning.

Reverse engineering (RE) is a domain of current interest where physical models are measured or digitized in order to obtain a virtual model. Currently, these virtual models are rebuilt using approaches that consider only a geometric point of view. These models are generally frozen (i.e., poorly parameterized and not easy to modify). These kinds of models are good for many applications but people need more for redesign operations. This paper proposes a knowledge-based approach for reverse engineering that enables a RE user to rebuild nonfrozen models that are close to an original CAD model.

This paper considers the problem of inferring the geometry of an object from values of the signed distance sampled on a uniform grid. The problem is motivated by the desire to effectively and efficiently model objects obtained by 3D imaging technology such as magnetic resonance, computed tomography, and positron emission tomography. Techniques recently developed for automated segmentation convert intensity to signed distance, and the voxel structure imposes the uniform sampling grid. The specification of the signed distance function (SDF) throughout the ambient space would provide an implicit and function-based representation (f-rep) model that uniquely specifies the object, and we refer to this particular f-rep as the signed distance function representation (SDF-rep). However, a set of uniformly sampled signed distance values may uniquely determine neither the distance function nor the shape of the object. Here, we employ essential properties of the signed distance to construct the upper and lower bounds on the allowed variation in signed distance, which combine to produce interval-valued extensions of the signed distance function. We employ an interval extension of the signed distance function as an interval SDF-rep that defines the range of object geometries that are consistent with the sampled SDF data. The particular interval extensions considered include a tight global extension and more computationally efficient local extensions that provide useful criteria for root exclusion/isolation. To illustrate a useful application of the interval bounds, we present a reliable approach to top-down octree membership classification for uniform samplings of signed distance functions.

In reverse engineering, 2D profile curve reconstruction based on cross-sectional points is a very crucial step for surface reconstruction such as lofting surface, swept surface, translational surface, and rotational surface. Unlike the traditional constrained fitting method that assumes that the cross-sectional points have been segmented in advance, and that the initial fitted curves are very close to the points, we propose a nonrigid registration method, through which a template curve can be automatically transformed and deformed to best fit the cross-sectional points. Compared with constrained fitting, nonrigid registration does not need any data preprocessing such as sorting, segmentation, and parametrization. The simulated and real examples have demonstrated the effectiveness and superiority of nonrigid registration for 2D blade profile curve reconstruction.

This paper presents an approach to automatically recover mesh surfaces with sharp edges for solids from their binary volumetric discretizations (i.e., voxel models). Our method consists of three steps. The topology singularity is first eliminated on the binary grids so that a topology correct mesh M 0 can be easily constructed. After that, the shape of M 0 is refined, and its connectivity is iteratively optimized into M n . The shape refinement is governed by the duplex distance fields derived from the input binary volume model. However, the refined mesh surface lacks sharp edges. Therefore, we employ an error-controlled variational shape approximation algorithm to segment M n into nearly planar patches and then recover sharp edges by applying a novel segmentation-enhanced bilateral filter to the surface. Using the technique presented in this paper, smooth regions and sharp edges can be automatically recovered from raw binary volume models without scalar field or Hermite data Compared to other related surface recovering methods on binary volume, our algorithm needs less heuristic coefficients.

In this paper we discuss the challenges of processing and converting 3D scanned data to representations suitable for interactive manipulation in the context of virtual restoration applications. We present a constrained parametrization approach that allows us to represent 3D scanned models as parametric surfaces defined over polyhedral domains. A combination of normal- and spatial-based clustering techniques is used to generate a partition of the model into regions suitable for parametrization. Constraints can be optionally imposed to enforce a strict correspondence between input and output features. We consider two types of virtual restoration methods: (a) a paint restoration method that takes advantage of the normal-based coarse partition to identify large regions of reduced metric distortion suitable for texture mapping and (b) a shape restoration approach that relies on a refined partition used to convert the input model to a multiresolution subdivision representation suitable for intuitive interactive manipulation during digital studies of historical artifacts.

An accurate three-dimensional (3D) mesh of biological models is fundamental for analysis and treatment simulations. Generally noninvasive magnetic resonance image (MRI) data are taken as the input for the simulation. The topologic relationship of anatomy is extracted from MR images through segmentation processes. To accelerate the biological modeling phase, template surface and volume meshes are generated based on MR images and∕or anatomical atlases (e.g., brain atlas, etc.). The boundary surfaces are extracted from segmented regions on the image slices, which are used as the input for 3D volume mesh generation. An intuitive graphic user interface was developed for biomedical applications. It integrated MRI data manipulation with surface mesh and volume mesh generators. Image volume and mesh geometries are registered in the MRI working space. As the core component of the system, a robust 3D mesh generation approach is presented. It is capable of describing irregular geometries exhibiting concave and convex surfaces. It uses deltahedral building blocks for volume mesh generation and creates high-quality, regular-shaped tetrahedral mesh elements. The approach supports multiple levels of localized refinement without reducing the overall mesh quality. The validity of this new mesh generation strategy and implementation is demonstrated via the medical applications in brain vasculature modeling, multimodality imaging for breast cancer detection, and numerous anatomically accurate models presented. Multiple material boundaries are preserved in each mesh with fidelity.

A segmentation and model-reconstruction algorithm is proposed based on polynomial approximation and on a new version of the “region growing” methodology. First, an initial partition is calculated on the basis of differential-geometric properties of the range image. Then, the first merging procedure is applied (“merge with constraints”) aiming at correctly identifying principal surfaces of the model. It examines all possible mergers of regions and selects those satisfying strict compatibility constraints. The second merging procedure relaxes these constraints to produce the “extended” regions and surfaces of the final segmentation. Theoretical work is presented proving the consistency of these merging procedures. Finally, application of the algorithm on industrial data is presented demonstrating the efficiency of the proposed methodology.

In this paper, a systematic scheme is proposed and novel technologies are developed to automatically reconstruct a CAD model from a set of point clouds scanned from the boundary surface of an existing object. The proposed scheme is composed of three major steps. In the first step, multiple input point clouds are incrementally integrated into a watertight triangle mesh to recover the object shape. In the second step, mesh segmentation is applied to the triangle mesh to extract individual geometric feature surfaces. Finally, the manifold topology describing the connectivity information between different geometric surfaces is automatically extracted and the mathematical description of each geometric feature is computed. The computed topology and geometry information represented in ACIS modeling kernel form a CAD model that may be used for various downstream applications. Compared with prior work, the proposed approach has the unique advantage that the processes of recognizing geometric features and of reconstructing CAD models are fully automated. Integrated with state of the art scanning devices, the developed model reconstruction method can be used to support reverse engineering of high precision mechanical components. It has potential applications to many engineering problems with a major impact on rapid design and prototyping, shape analysis, and virtual reality.