Search

Fringe profilometry is a method that obtains the 3D information of objects by projecting a pattern of fringes. The three-step technique uses only three images to acquire the 3D information from an object, and many studies have been conducted to improve this technique. However, there is a problem that is inherent to this technique, and that is the quasi-periodic noise that appears due to this technique and considerably affects the final 3D object reconstructed. Many studies have been carried out to tackle this problem to obtain a 3D object close to the original one. The application of deep learning in many areas of research presents a great opportunity to to reduce or eliminate the quasi-periodic noise that affects images. Therefore, a model of convolutional neural network along with four different patterns of frequencies projected in the three-step technique is researched in this work. The inferences produced by models trained with different frequencies are compared with the original ones both qualitatively and quantitatively. [ABSTRACT FROM AUTHOR]

The modeling of real objects digitally is an area that has generated a high demand due to the need to obtain systems that are able to reproduce 3D objects from real objects. To this end, several techniques have been proposed to model objects in a computer, with the fringe profilometry technique being the one that has been most researched. However, this technique has the disadvantage of generating Moire noise that ends up affecting the accuracy of the final 3D reconstructed object. In order to try to obtain 3D objects as close as possible to the original object, different techniques have been developed to attenuate the quasi/periodic noise, namely the application of convolutional neural networks (CNNs), a method that has been recently applied for restoration and reduction and/or elimination of noise in images applied as a pre-processing in the generation of 3D objects. For this purpose, this work is carried out to attenuate the quasi/periodic noise in images acquired by the fringe profilometry technique, using a modified CNN-Multiresolution network. The results obtained are compared with the original CNN-Multiresolution network, the UNet network, and the FCN32s network and a quantitative comparison is made using the Image Mean Square Error E (IMMS), Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and Profile (MSE) metrics. [ABSTRACT FROM AUTHOR]

Fringe profilometry is a method that obtains the 3D information of objects by projecting a pattern of fringes. The three-step technique uses only three images to acquire the 3D information from an object, and many studies have been conducted to improve this technique. However, there is a problem that is inherent to this technique, and that is the quasi-periodic noise that appears due to this technique and considerably affects the final 3D object reconstructed. Many studies have been carried out to tackle this problem to obtain a 3D object close to the original one. The application of deep learning in many areas of research presents a great opportunity to to reduce or eliminate the quasi-periodic noise that affects images. Therefore, a model of convolutional neural network along with four different patterns of frequencies projected in the three-step technique is researched in this work. The inferences produced by models trained with different frequencies are compared with the original ones both qualitatively and quantitatively.

This paper consists of a bibliometric study that covers the topic of 3D object detection from 2022 until the present day. It employs various analysis approaches that shed light on the leading authors, affiliations, and countries within this research domain alongside the main themes of interest related to it. The findings revealed that China is the leading country in this domain given the fact that it is responsible for most of the scientific literature as well as being a host for the most productive universities and authors in terms of the number of publications. China is also responsible for initiating a significant number of collaborations with various nations around the world. The most basic theme related to this field is deep learning, along with autonomous driving, point cloud, robotics, and LiDAR. The work also includes an in-depth review that underlines some of the latest frameworks that took on various challenges regarding this topic, the improvement of object detection from point clouds, and training end-to-end fusion methods using both camera and LiDAR sensors, to name a few. [ABSTRACT FROM AUTHOR]

The fast advancement of science and technology has transformed product designing by allowing designers to produce high-quality goods that fulfill the needs of customers and industries. With its realistic visual effects and immersive experience, virtual reality (VR) offers a new method of product designing by replacing expensive and inflexible physical prototypes. This research paper discusses an evaluation system for virtual product designing using VR-assisted approach to overcome these issues. The proposed approach aims to offer a quick and low-cost method for reviewing and analyzing product ideas. It uses a product display system for professionals to evaluate the designs and allows the experts to assess the product programs using evaluation indexes. The analysis procedure is based on evaluations of user experiences, which are divided into three categories: behavior layer experience, sensor layer experience, and reflection layer experience. The evaluation index structure is built using a hierarchical inference approach. The experimental findings performed in the study demonstrate the superiority of the proposed algorithm and system over existing algorithms and systems. The proposed algorithm regularly obtains higher prediction values by demonstrating improved predictive accuracy for athletic performance. Furthermore, the proposed system has increased variety, significant variances, and statistical significance when evaluating product design, resulting in higher customer satisfaction. These findings illustrate potential of the suggested approach to improve product design and overall consumer pleasure by providing significant insights for the engineering and industrial sectors. [ABSTRACT FROM AUTHOR]

The modeling of real objects digitally is an area that has generated a high demand due to the need to obtain systems that are able to reproduce 3D objects from real objects. To this end, several techniques have been proposed to model objects in a computer, with the fringe profilometry technique being the one that has been most researched. However, this technique has the disadvantage of generating Moire noise that ends up affecting the accuracy of the final 3D reconstructed object. In order to try to obtain 3D objects as close as possible to the original object, different techniques have been developed to attenuate the quasi/periodic noise, namely the application of convolutional neural networks (CNNs), a method that has been recently applied for restoration and reduction and/or elimination of noise in images applied as a pre-processing in the generation of 3D objects. For this purpose, this work is carried out to attenuate the quasi/periodic noise in images acquired by the fringe profilometry technique, using a modified CNN-Multiresolution network. The results obtained are compared with the original CNN-Multiresolution network, the UNet network, and the FCN32s network and a quantitative comparison is made using the Image Mean Square Error E (IMMS), Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and Profile (MSE) metrics.

Robot-guided inkjet printing technology offers a new way for the digital and additive deposition of low-viscous inks to be made directly onto arbitrary surfaces and, thus, enables the production of individualized printed electronics on large-scale objects. When compared to conventional flatbed printing, the distance between the nozzle plate and the object's surface varies and needs to be considered in order to match the accuracy requirements needed for the positioning of single drops. Knowledge about applicable distance limits and the influence of tunable print parameters is crucial for improving the print process and results. This study discusses the sources of errors in the inkjet printing process onto 3D objects and presents extensive results about position accuracy in relation to jetting distance for different parameter sets of functional inks, drop volumes, and piezo voltages. Additionally, an efficient novel method was applied to determine the drop position accuracy of inkjet droplets in relation to the jetting distance. The method relies on cylinder geometry for the object and an inkjet head that is guided by a six-axis robot manipulator along the cylinder's axis. For the determination of drop placement accuracy, the position of single dots on the surface was compared to a model which considered the cylinder radii, drop velocity, and the movement speed of the guided inkjet printhead. The method and the extensive research results can be utilized for the prediction of achievable drop placement accuracy and the prior definition of distance limits. [ABSTRACT FROM AUTHOR]

The current color-difference formulas were developed based on 2D samples and there is no standard guidance for the color-difference evaluation of 3D objects. The aim of this study was to test and optimize the CIELAB and CIEDE2000 color-difference formulas by using 42 pairs of 3D-printed spherical samples in Experiment I and 40 sample pairs in Experiment II. Fifteen human observers with normal color vision were invited to attend the visual experiments under simulated D65 illumination and assess the color differences of the 82 pairs of 3D spherical samples using the gray-scale method. The performances of the CIELAB and CIEDE2000 formulas were quantified by the STRESS index and F-test with respect to the collected visual results and three different optimization methods were performed on the original color-difference formulas by using the data from the 42 sample pairs in Experiment I. It was found that the optimum parametric factors for CIELAB were k L = 1.4 and k C = 1.9, whereas for CIEDE2000, k L = 1.5. The visual data of the 40 sample pairs in Experiment II were used to test the performance of the optimized formulas and the STRESS values obtained for CIELAB/CIEDE2000 were 32.8/32.9 for the original formulas and 25.3/25.4 for the optimized formulas. The F-test results indicated that a significant improvement was achieved using the proposed optimization of the parametric factors applied to both color-difference formulas for 3D-printed spherical samples. [ABSTRACT FROM AUTHOR]

There are various structures in the underground space, and the structures are managed as 3D data in underground geospatial information map system (UGIMS). However, there is a problem that takes a long time to load the data into a mobile app from UGIMS because the data are very large. In order to reduce the loading time of 3D data, the number of data vertices should be reduced within a range that can keep the shape of 3D object in a mobile app. In this paper, we try to reduce the number of vertices using a 3D Mesh Simplification method to reduce the loading time of data in a mobile app. The proposed method applies simplification by extracting refined vertex feature information using a deep learning encoder‐decoder based model and performing vertex clustering by grouping similar vertices together with the feature information using K‐means clustering. We were able to reduce the number of vertices by 30% compared with the previous data. © 2022 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC. [ABSTRACT FROM AUTHOR]

Since the number of 3D models is rapidly increasing, extracting better feature descriptors to represent 3D models is very challenging for effective 3D model retrieval. There are some problems in existing 3D model representation approaches. For example, many of them focus on the direct extraction of features or transforming 3D models into 2D images for feature extraction, which cannot effectively represent 3D models. In this paper, we propose a novel 3D model feature representation method that is a kind of voxelization method. It is based on the space-based concept, namely CSS (Cube of Space Sampling). The CSS method uses cube space 3D model sampling to extract global and local features of 3D models. The experiments using the ESB dataset show that the proposed method to extract the voxel-based features can provide better classification accuracy than SVM and comparable retrieval results using the state-of-the-art 3D model feature representation method. [ABSTRACT FROM AUTHOR]

The rapid development of computer vision has led to an increasing amount of 3D data, such as multiple views and point clouds, which are widely used in 3D object recognition and retrieval. Intuitively, the quality of 3D data is the most crucial factor that directly affects the performance of 3D applications. However, how to evaluate the 3D data quality, especially the multi-view data quality, is still an open question. To tackle this issue, we propose an entropy-based multi-view information quantification model (MV-Info model) to quantitatively evaluate the multi-view data information. Our proposed MV-Info model consists of hierarchical data module, feature generation module, and quantitative calculation module. Besides, it considers the information entropy theory for more reasonable quantification results. In our method, how much information we can observe from a group of views can be quantified, which can be used to support 3D recognition and retrieval. We also designed a series of experiments to evaluate the effectiveness of the proposed model. The experimental results demonstrate the rationality and validity of the proposed model. [ABSTRACT FROM AUTHOR]

By means of a case study, the submitted article describes the solution of a single part of a larger research task. In the submitted article, we deal with the process of creating 3D models of three selected cultural heritage objects (ceramic vase of Pozdišovce, traditional mangle board and traditional music instrument - jaw harp) using two different methods - manual modelling and photogrammetry. As a conclusion, we need to find the method, the use of which leads to better and higher quality results, taking into account the time consumption and other factors. We will use the created virtual 3D equivalents of real objects later in the next phase of research when they will become part of a small test scene of a „virtual museum“. In the museum, the final 3D objects will be presented to the participants in the following phase of our research. [ABSTRACT FROM AUTHOR]

To accurately recognize and locate the surrounding vehicles and pedestrians, an attention-based PointPillars+ 3D target detection algorithm was proposed. The entire space was uniformly divided into pillars with a given resolution, and a pseudo-image was generated by extracting point-based features from all pillars. Two attention modules were introduced to highlight and restrain the information in the pseudo-image. A convolution neural network was used to process the output of the attention module, and the single shot multibox detector（SSD） was used for 3D object detection. The evaluation results show that the parallel attention-based PointPillars achieves good performance. Compared with the traditional PointPillars, the mAPm is increased from 66.19 to 69.95 in the bird′s eye view， and the car mAP is increased from 86.10 to 87.73. In the 3D mode, the mAPm is increased from 59.20 to 62.55, and the car mAP is increased from 74.99 to 76.25. [ABSTRACT FROM AUTHOR]

Watermarking provides an efficient mechanism for copyright protection of digital media by embedding information in the object to identify the owner. This paper proposed a watermarking technique that can successfully embed a grey-scale image in a 3D mesh object. In our proposed technique, each three vertices of the 3D object can hide one pixel from the secret image. The experimental results showed that the proposed technique has better visual quality on both the watermarked object and the extracted secret image. [ABSTRACT FROM AUTHOR]

Since the number of 3D models is rapidly increasing, extracting better feature descriptors to represent 3D models is very challenging for effective 3D model retrieval. There are some problems in existing 3D model representation approaches. For example, many of them focus on the direct extraction of features or transforming 3D models into 2D images for feature extraction, which cannot effectively represent 3D models. In this paper, we propose a novel 3D model feature representation method that is a kind of voxelization method. It is based on the space-based concept, namely CSS (Cube of Space Sampling). The CSS method uses cube space 3D model sampling to extract global and local features of 3D models. The experiments using the ESB dataset show that the proposed method to extract the voxel-based features can provide better classification accuracy than SVM and comparable retrieval results using the state-of-the-art 3D model feature representation method.

Deep learning approaches to estimating full 3D orientations of objects, in addition to object classes, are limited in their accuracies, due to the difficulty in learning the continuous nature of three-axis orientation variations by regression or classification with sufficient generalization. This paper presents a novel progressive deep learning framework, herein referred to as 3D POCO Net, that offers high accuracy in estimating orientations about three rotational axes yet with efficiency in network complexity. The proposed 3D POCO Net is configured, using four PointNet-based networks for independently representing the object class and three individual axes of rotations. The four independent networks are linked by in-between association subnetworks that are trained to progressively map the global features learned by individual networks one after another for fine-tuning the independent networks. In 3D POCO Net, high accuracy is achieved by combining a high precision classification based on a large number of orientation classes with a regression based on a weighted sum of classification outputs, while high efficiency is maintained by a progressive framework by which a large number of orientation classes are grouped into independent networks linked by association subnetworks. We implemented 3D POCO Net for full three-axis orientation variations and trained it with about 146 million orientation variations augmented from the ModelNet10 dataset. The testing results show that we can achieve an orientation regression error of about 2.5° with about 90% accuracy in object classification for general three-axis orientation estimation and object classification. Furthermore, we demonstrate that a pre-trained 3D POCO Net can serve as an orientation representation platform based on which orientations as well as object classes of partial point clouds from occluded objects are learned in the form of transfer learning.

This paper presents a new method of secret three-dimensional object sharing (S3DOS), which allows sharing of three-dimensional (3-D) objects, while preserving its file format by selectively encrypting a 3-D object in order to sufficiently protect the visual nature of the content. This scheme, named format-compliant selective (k, n) S3DOS, modifies selected bits of the vertices of a 3-D object to protect the visual content by inducing geometrical distortions. These distortions are controlled thanks to the application of a degradation level at the beginning of the sharing process. To reconstruct the secret 3-D object, at least $k$ shared 3-D objects among $n$ have to be combined in order to remove the geometrical distortions and recover the exact original 3-D object. Experimental results are presented and evaluated to showcase the feasibility of the proposed scheme. [ABSTRACT FROM AUTHOR]

Abstract Archaeozoology or zooarchaeology is the scientific field that aims to rebuild natural and cultural relations between human and animals. To reach that objective, archaeozoologists study faunal remains, such as bones. In this regard, a 3D representation is put into place. This would allow a better understanding of the initial spatial positions of the animals in the pits, and would make it possible to examine in a more rigorous manner the explanatory hypotheses by manipulating the 3D view of the pit and bones. This manipulation invokes the use of computer vision techniques, including 3D object recognition methods. In this work we present a novel approach to recognize 3D bones in occluded and cluttered scenes inspiring from a stat of the art method called spin images. We combine this method with saliency concept to improve the the robustness to occlusions and clutters significantly and also to decrease the complexity of the algorithm. [ABSTRACT FROM AUTHOR]

Human body anatomy becomes an important topic in Biology subject that must be understood since junior high school. Learning materials are mostly available in form of book and anatomy mannequin (puppet), but it is still insufficient enough to help students in understanding human body anatomy. Augmented Reality (AR) is a technology that combines a real thing into virtual environment interactively. This research purpose is to develop an AR application for human body anatomy learning to be more interesting and easier for student to understand. This application enables student to learn human body anatomy with 3D object interaction while previously using textbook and mannequin. Research method in for this study is by using quantitative method that collects data and then develops the prototype to prove the impact. Application development method is done by using waterfall method that includes planning (collect data and analysis), design (user interface and diagram), implementation, and testing. Research result is AR application for human body anatomy learning that contains 3D object, organ explanation and position that can be accessible on web. [ABSTRACT FROM AUTHOR]

In this paper, we exploit the potential of the GPU in order to accelerate the process of 3D shape retrieval in large databases. Indeed, the massive parallelism of the GPU offers a huge performance in much high-performance computing (HPC) applications. Our solution consists to accelerate the shape matching process of methods that use a specific similarity metric called Clock Matching (CM). This CM measure is used by view-based methods as an efficient solution to compare two 3D models even if they are not presented in same pose and orientation by taking into account all possible poses in the matching phase. However, the increase in the number of comparisons has a strong influence on the execution time. Our challenge is to exploit the maximum benefit of GPU computing resource by considering the difficulty of implementing the CM metric on GPU. Indeed, the descriptor of a given 3D object is organized using a specific data structure (hash table), where only the information whose values are not equal to zero appears in the feature vector, which makes the parallelization on GPU to be not trivial. Experiment results show a reasonable benefit from the GPU approach. [ABSTRACT FROM AUTHOR]

Grasp planning is crucial for many robotic applications such as object manipulation and object transport. Planning stable grasps is a challenging problem. Many parameters such as object geometry, hand geometry and kinematics, hand-object contacts have to be considered, making the space of grasps too large to be exhaustively searched. This paper presents a general approach for planning grasps on 3D objects based on hand-object geometric fitting. Our key idea is to build a contact score map on a 3D object’s voxelization, and apply this score map and a hand’s kinematic parameters to find a set of target contacts on the object surface. Guided by these target contacts, we find grasps with a high quality measure by iteratively adjusting the hand pose and joint angles to fit the hand’s instantaneous geometric shape with the object’s fixed shape, during which the fitting process is speeded up by taking advantage of the discrete volumetric space. We demonstrate the effectiveness of our grasp planning approach on 3D objects of various shapes, poses, and sizes, as well as hand models with different kinematics. A comparison with two state-of-the-art approaches shows that our approach can generate grasps that are more likely to be stable, especially for objects with complex shapes. [ABSTRACT FROM AUTHOR]

Dewata Nawa Sanga in Hinduism has an important role as the foundation of the religious concept implemented by Hindus in Bali, Indonesia. It has the meaning of nine deities or manifestations of Ida Sang Hyang Widhi Wasa (God Almighty) that guard or control the nine wind direction. Dewata Nawa Sanga in terms of education in Bali has been learned by the elementary school students. However, in social life the concept of Dewata Nawa Sanga itself has not been attached and understood by all Balinese Hinduism yet. Based on the problem in social life, there is an idea to develop Markerless Augmented Reality Utilizing Gyroscope to Demonstrate the Position of Dewata Nawa Sanga. This application is designed to know, recognize, and understand the characteristics of Dewata Nawa Sanga using gyroscope. This sensor serves to fulfill the object of the deities in the coordinates to be determined. Furthermore, it serves to provide information about Dewata Nawa Sanga along with the innovative and informative 3D animation. The result of usefulness questionnaire of this application percentage value average is 88.4%. It can be concluded that AR Dewata Nawa Sanga is very useful and helpful for learners in exploring the Dewata Nawa Sanga position in a real-world environment. The result of usability and satisfaction questionnaire of this application percentage value average is 84.8%. It shows that the application is very useful for the participants to learn Dewata Nawa Sanga and very satisfied to use. [ABSTRACT FROM AUTHOR]

Every year the edge between the real and digital worlds is becoming more and more blurred. Augmented and virtual reality rapid development creates new opportunities for more productive work and entertainment, revolution in 3D printing technologies begets boost in multiple DIY communities appearance and sharing economy growth. All these factors require new technologies that allow making 3D models from real world objects, but most of these solutions are either very expensive or require complex technical knowledge that most ordinary people do not have. This paper provides a review and comparison of modern methods for 3D models of physical objects real time reconstruction that can be used in present-day mobile solutions.

Company X is a reputable company that has more than 1000 people employees. One of the routines that performed by company X is orientation for new employees, and there is an event that invites them to visit the factory. The location of the main office and the factory is far, so this activity lacks time efficiency. This application aims to facilitate company X in conducting new employee orientation so that the location can be seen in the virtual world. The first thing in the process of making the application is visiting the factory to determine the shape and position of the object, so that the expected results are shown to resemble its original condition. 3d object and its texture are created using Blender application that can be imported into the made application. Collision feature aims to prevent users from penetrating the objects.The position, shape and location of objects contained in the application already resemble the original plant site. The main obstacle in making this application is the creation time, so the detailed map of the plant cannot be completed. The main building, the production building and the warehouse have been made, only small objects contained in the main building have not been contained.

Every year the edge between the real and digital worlds is becoming more and more blurred. Augmented and virtual reality rapid development creates new opportunities for more productive work and entertainment, revolution in 3D printing technologies begets boost in multiple DIY communities appearance and sharing economy growth. All these factors require new technologies that allow making 3D models from real world objects, but most of these solutions are either very expensive or require complex technical knowledge that most ordinary people do not have. This paper provides a review and comparison of modern methods for 3D models of physical objects real time reconstruction that can be used in present-day mobile solutions. [ABSTRACT FROM AUTHOR]

Context Cardiovascular disease remains the leading cause of death affecting the adult population. It involves disorders of myocardial perfusion which can lead to the disturbance in myocardial function and/or myocardial infarction. Finding an efficient method for the deformation analysis in the left ventricle of the heart (LV) is one of the major concerns in cardiac imaging. Objective In literature, most models provide global representation of the LV surface, that cannot specify the affected area. The present work is aimed at quantifying the local deformations in the LV by developing accurate computational approaches in order to refine the diagnosis and specify the ischemic territory. Method The proposed approach advocates the application of three different methods: (1) the numerical quantification of regional deformation based on SPHARM shape descriptors, (2) volume evolution, and (3) the Hotelling metric computation by use of the relative curvature at the vertices of the triangulation. In order to progress to a regional analysis, we carried out a division into 17 regions according to the AHA standard of LV object. It is worth noting that our two first methods are based on the application of the cloud closing process by generating new points on the cutting planes. Results The different approaches were tested on 44 patients. The obtained results show its effectiveness and validity by giving the degree and the location of the pathological deformation in the LV. We compared the obtained results with diagnostic offered by qualified nuclear medicine physicians. Also, we compared the obtained results of each methods in order to define the more robust approach. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel geometric features based method to categorize 3D models using probabilistic neural network and support vector machine classifiers. The employed features are extracted from face and vertex characteristics. In addition, we utilize the proposed features in 3D object retrieval. To achieve this end, each model is decomposed into a set of local/global geometrical features. We use histograms of two variables, i.e., deviation angle of normal vector on the object surface point from the vector that connect shape center to that point; and distance of object surface point from shape center. To achieve better separability of different models, mutual Euclidean distance histogram for the pairs of surface points is also used. The most advantage of using histogram to represent the features is that it shows the density of data and enables creating of low dimensional feature vector and consequently decreasing of computational cost in classification process. The effectiveness of our proposed 3D object categorization system has been evaluated on the generalized McGill 3D model dataset in terms of both accuracy and speed measures. Widespread experimental results and comparison with the other similar methods, demonstrate efficiency of the proposed approach to improve both accuracy and speed of categorization system.

Calculations of a dedicated optical system for the quality inspection of cylindrical surfaces are presented. Peculiarities of imaging a cylindrical surface by means of specially enhanced inherent field curvature are considered. Results of using the system for the inspection of quality of fuel pellets are provided. [ABSTRACT FROM AUTHOR]