Large-scale 3D Research Articles

Hierarchical Structures Computational Design and Digital 3D Printing

Current advances in construction automation, especially in large-scale additive manufacturing, highlight the vast potential for robots in architecture. Robotic construction is unique in its potential to reproduce highly complex structures. To advance the question of how rapid prototyping techniques are adopted in large-scale 3D printing of forms and structures, this paper presents computational methods for the design and robotic construction of cellular membranes. This research presents a comprehensive morphological model of structurally differentiated cellular membranes based on the theoretical biology model of hierarchical structures found in natural cellular solids, and, more specifically, in trabecular bone. The morphological model originates from a system of forces in equilibrium; therefore, it presents the geometric homology of a static tensional system. This research offers a methodology for the design and manufacture of meso- to large-scale triangulated geometric configurations by discrete design methods that are suitable for the robotic fused deposition of lattices and their architectural implementation in the automated manufacturing of shell structures. First, this paper explores how a form can be digitally created by geometrically emulating a given static system of forces in space. Second, inspired by the complex mechanical behavior of cancellous bone, we apply hierarchical principles found in bone remodeling to characterize discrete units that conform to continuous trabecular-like lattices. We study the geometry, limitations, opportunities for optimization, and mechanical characteristics of the lattice. The computational design methods and additive manufacturing techniques are tested in the design and construction hierarchical structures.

Large-scale 3D object segmentation employing conditional angular clustering technique

This paper aims to develop an automatic 3D object segmentation method for the large-scale point clouds. Given a range image, the preprocessing is first applied to get the optimal 3D point cloud. A [Formula: see text]-nearest neighbor is built, and a segmentation algorithm based on the conditional angular clustering technique is used to segment the objects from the point cloud. The algorithm is tested on the real point cloud datasets. The experiment results demonstrated that the developed segmentation method can be used to localize the object with the relative uncertainty of 0.27%.

Non-planar multinozzle additive manufacturing of thermoset composite microscaffold networks

Material extrusion-based non-planar additive manufacturing (AM) using 6-axis industrial robot arms shows great promise for three-dimensional (3D) printing of non-planar parts, despite its typical use of a single large coarse nozzle and high printing time. Here, a non-planar multinozzle AM method capable of achieving high resolution large-scale 3D printing with short printing time using 6-axis robots is reported. A custom Multinozzle Toolpath Generator (MTG), based on non-planar parameterized surface projection, is used to compute the orientation of a multinozzle extrusion array along the curvature of a 3D scanned substrate. The method is demonstrated by the fast 3D printing of large periodic structures (up to an area of ∼5.2 × 104 mm2 within 20 min) consisting of interconnected epoxy composite microscaffolds on a single curvature sinusoidal surface and a double curvature surface of an actual aircraft casing, with an error of 15 μm. The method could be used for AM in various fields such as tissue engineering, microfluidic channels, and porous acoustic coatings of aircraft components.

A Cluster-Based 3D Reconstruction System for Large-Scale Scenes

The reconstruction of realistic large-scale 3D scene models using aerial images or videos has significant applications in smart cities, surveying and mapping, the military and other fields. In the current state-of-the-art 3D-reconstruction pipeline, the massive scale of the scene and the enormous amount of input data are still considerable obstacles to the rapid reconstruction of large-scale 3D scene models. In this paper, we develop a professional system for large-scale 3D reconstruction. First, in the sparse point-cloud reconstruction stage, the computed matching relationships are used as the initial camera graph and divided into multiple subgraphs by a clustering algorithm. Multiple computational nodes execute the local structure-from-motion (SFM) technique, and local cameras are registered. Global camera alignment is achieved by integrating and optimizing all local camera poses. Second, in the dense point-cloud reconstruction stage, the adjacency information is decoupled from the pixel level by red-and-black checkerboard grid sampling. The optimal depth value is obtained using normalized cross-correlation (NCC). Additionally, during the mesh-reconstruction stage, feature-preserving mesh simplification, Laplace mesh-smoothing and mesh-detail-recovery methods are used to improve the quality of the mesh model. Finally, the above algorithms are integrated into our large-scale 3D-reconstruction system. Experiments show that the system can effectively improve the reconstruction speed of large-scale 3D scenes.

Epitaxial Growth of Large Area Two-Dimensional Ferroelectric α-In2Se3.

Two-dimensional (2D) ferroelectric materials have attracted intensive attention in recent years for academic research. However, the synthesis of large-scale 2D ferroelectric materials for electronic applications is still challenging. Here, we report the successful synthesis of centimeter-scale ferroelectric In2Se3 films by selenization of In2O3 in a confined space chemical vapor deposition method. The as-grown homogeneous thin film has a uniform thickness of 5 nm with robust out-of-plane ferroelectricity at room temperature. Scanning transmission electron microscopy and Raman spectroscopy reveal that the thin film is 2H stacking α-In2Se3 with excellent crystalline quality. Electronic transport measurements of In2Se3 highlight the current-voltage hysteresis and polarization modulated diode effect due to the switchable Schottky barrier height (SBH). First-principles calculations reveal that the polarization modulated SBH is originated from the competition between interface charge transfer and polarized charge. The large area growth of epitaxial In2Se3 opens up potential applications of In2Se3 in novel nanoelectronics.

Data Preparation Impact on Semantic Segmentation of 3D Mobile LiDAR Point Clouds Using Deep Neural Networks

Currently, 3D point clouds are being used widely due to their reliability in presenting 3D objects and accurately localizing them. However, raw point clouds are unstructured and do not contain semantic information about the objects. Recently, dedicated deep neural networks have been proposed for the semantic segmentation of 3D point clouds. The focus has been put on the architecture of the network, while the performance of some networks, such as Kernel Point Convolution (KPConv), shows that the way data are presented at the input of the network is also important. Few prior works have studied the impact of using data preparation on the performance of deep neural networks. Therefore, our goal was to address this issue. We propose two novel data preparation methods that are compatible with typical density variations in outdoor 3D LiDAR point clouds. We also investigated two already existing data preparation methods to show their impact on deep neural networks. We compared the four methods with a baseline method based on point cloud partitioning in PointNet++. We experimented with two deep neural networks: PointNet++ and KPConv. The results showed that using any of the proposed data preparation methods improved the performance of both networks by a tangible margin compared to the baseline. The two proposed novel data preparation methods achieved the best results among the investigated methods for both networks. We noticed that, for datasets containing many classes with widely varying sizes, the KNN-based data preparation offered superior performance compared to the Fixed Radius (FR) method. Moreover, this research allowed identifying guidelines to select meaningful downsampling and partitioning of large-scale outdoor 3D LiDAR point clouds at the input of deep neural networks.

A DISTRIBUTED BUNDLE ADJUSTMENT METHOD USING ITERATIVE MOTION AVERAGING

Abstract. With the wide use of UAVs in urban scenes making image acquisition easier, processing large-scale 3D reconstruction problems have become a basic need of the Structure from Motion (SfM) community. Because of high memory consumption and computation complexity, bundle adjustment (BA) is a significant bottleneck in large-scale 3D reconstruction problems. Distributed strategy is a current research direction. Recent research cut the view graph to optimize sub-scenes in small size and average common cameras and/or 3D points to align them into one in the end. Yet, those methods tried to simply average cameras and/or 3D points, which lack mathematical rigor and tend to get a sub-optimal result. This paper first cuts the view graph and expands it. Then we use the strategy of motion averaging instead of simply averaging points to obtain a robust result. Finally, refined results are achieved by optimizing reprojection errors in the overlapping area. We conduct experiments on UAV data, check the validity of graph cut and expansion, and analyze whether graph size influences the precision computation time. Visualization results show that our method preserves scenes well; precision and visualization of un-expanded scenes prove the necessity of graph expansion; statistical results also indicate the negative infect of large sub-scene size.

Identification of Promising Drug Candidates against Prostate Cancer through Computationally-Driven Drug Repurposing.

Prostate cancer (PC) is one of the most common types of cancer in males. Although early stages of PC are generally associated with favorable outcomes, advanced phases of the disease present a significantly poorer prognosis. Moreover, currently available therapeutic options for the treatment of PC are still limited, being mainly focused on androgen deprivation therapies and being characterized by low efficacy in patients. As a consequence, there is a pressing need to identify alternative and more effective therapeutics. In this study, we performed large-scale 2D and 3D similarity analyses between compounds reported in the DrugBank database and ChEMBL molecules with reported anti-proliferative activity on various PC cell lines. The analyses included also the identification of biological targets of ligands with potent activity on PC cells, as well as investigations on the activity annotations and clinical data associated with the more relevant compounds emerging from the ligand-based similarity results. The results led to the prioritization of a set of drugs and/or clinically tested candidates potentially useful in drug repurposing against PC.

Class-aware tiny object recognition over large-scale 3D point clouds

Although tremendous strides have been made in object recognition over large-scale 3D point clouds, one of the remaining open challenges is detecting tiny objects. The signal and appearance information of tiny objects are generally sparse and insufficient due to their tiny size, which makes the tiny object difficult to be detected with existing studies both in tiny 2D object recognition and general 3D object recognition since 3D topology relations are lost due to feature projection in tiny 2D object recognition and massive background may dominate the feature learner in general 3D object recognition. To the best of our knowledge, we explore tiny object recognition for the first time over large-scale 3D point clouds by designing deep neural network. To cope with this problem, a novel two-stage approach is proposed in this paper, which can effectively learn an informative and discriminative feature representation and efficiently process large-scale point clouds. Specifically, a class-aware filtering strategy is designed to dispense with the redundant background information. After that, we introduce a novel aggregating and sampling scheme in a supervised manner to progressively increase the receptive field for each 3D point, which can adaptively capture useful signals from tiny objects and learn complex geometric structures. Last, we propose a biased loss function that is inversely correlated to the number of tiny object points, thereby effectively handling the sparsity of tiny objects. Extensive experimental results over two real-world data sets validate the effectiveness and efficiency of the proposed approach.

The electron diffusion region dominated by electromagnetic turbulence in the reconnection current layer

Most of the plasma fluid equations have employed the electrical resistivity to generate the magnetic dissipation required for magnetic reconnection to occur in collisionless plasma. However, there has been no clear evidence that such a model is indeed appropriate in the reconnection diffusion region in terms of the kinetic physics. The present study demonstrates that, using a large-scale 3D kinetic simulation and analytical analysis, the spatial distribution of the non-ideal electric field is consistent with the dissipation due to the viscosity rather than the resistivity, when electromagnetic (EM) turbulence is dominant in the electron diffusion region (EDR). The effective viscosity is caused by the EM turbulence that is driven by the flow shear instabilities leading to the electron momentum transport across the EDR. The result suggests a fundamental modification of the fluid equations using the resistivity in the Ohm's law. In contrast, for the 2D current sheet without significant turbulence activity, the non-ideal field profile does not obey the simple form based on the viscosity, so that further investigation is needed for a better description.

Advancing 3D medical image analysis with variable dimension transform based supervised 3D pre-training

The difficulties in both data acquisition and annotation substantially restrict the sample sizes of training datasets for 3D medical imaging applications. Therefore, it is non-trivial to build well-performing 3D convolutional neural networks from scratch. Previous efforts on 3D pre-training have frequently relied on self-supervised approaches, which use either predictive or contrastive learning on unlabeled data to build invariant 3D representations. However, because of the unavailability of large-scale supervision information, obtaining semantically invariant and discriminative representations from these learning frameworks remains problematic. In this paper, we revisit an innovative yet simple fully-supervised 3D network pre-training framework to take advantage of semantic supervision from large-scale 2D natural image datasets. With a redesigned 3D network architecture, reformulated natural images are used to address the problem of data scarcity and develop powerful 3D representations. Comprehensive experiments on five benchmark datasets demonstrate that the proposed pre-trained models can effectively accelerate convergence while also improving accuracy for a variety of 3D medical imaging tasks such as classification, segmentation, and detection. In addition, as compared to training from scratch, it can save up to 60% of annotation efforts. On the NIH DeepLesion dataset, it also achieves state-of-the-art detection performance, outperforming earlier self-supervised and fully-supervised pre-training approaches, as well as methods that do training from scratch. To facilitate further development of 3D medical models, our code and pre-trained model weights are publicly available at https://github.com/urmagicsmine/CSPR.

Visualized Study on a New Preformed Particle Gels (PPG) + Polymer System to Enhance Oil Recovery by Oil Saturation Monitoring Online Flooding Experiment

After tertiary recovery from the oilfields, improving the production of the remaining hydrocarbon is always challenging. To significantly improve oil recovery, a heterogeneous composite flooding system has been developed with preformed particle gels (PPG) and polymers according to the technical approach of plugging and flooding combination. In addition, an oil saturation monitoring device and a large-scale 3D physical model were designed to better evaluate the performance of the technique. The evaluation results show that the viscosity, stability, and elasticity of the heterogeneous composite flooding system are better than the single polymer system. In addition, both systems exhibit pseudoplastic fluid characteristics and follow the principle of shear thinning. The results of seepage experiments showed that PPG migrates alternately in porous media in the manner of "piling plugging-pressure increasing-deformation migration". The heterogeneous composite system can migrate to the depths of the oil layer, which improves the injection profile. In the visualization experiment, the heterogeneous composite system preferentially flowed into the high-permeability layer, which increased the seepage resistance and forced the subsequent fluid to flow into the medium and low permeability layers. The average saturation of the high, medium, and low permeability layers decreased by 4.74%, 9.51%, and 17.12%, respectively, and the recovery factor was further improved by 13.56% after the polymer flooding.

A Geometric Multigrid Method for 3D Magnetotelluric Forward Modeling Using Finite-Element Method

The traditional three-dimensional (3D) magnetotelluric (MT) forward modeling using Krylov subspace algorithms has the problem of low modeling efficiency. To improve the computational efficiency of 3D MT forward modeling, we present a novel geometric multigrid algorithm for the finite element method. We use the vector finite element to discretize Maxwell’s equations in the frequency domain and apply the Dirichlet boundary conditions to obtain large sparse complex linear equations for the solution of EM responses. To improve the convergence of the solution at low frequencies we use the divergence correction to correct the electric field. Then, we develop a V-cycle geometric multigrid algorithm to solve the linear equations system. To demonstrate the efficiency and effectiveness of our geometric multigrid method, we take three synthetic models (COMMEMI 3D-2 model, Dublin test model 1, modified SEG/EAEG salt dome model) and compare our results with the published ones. Numerical results show that the geometric multigrid algorithm proposed in this paper is much better than the commonly used Krylov subspace algorithms (such as SOR-GMRES, ILU-BICGSTAB, SOR-BICGSTAB) in terms of the iteration number, the solution time, and the stability, and thus is more suitable for large-scale 3D MT forward modeling.

Large-scale metagrating complex-based light field 3D display with space-variant resolution for non-uniform distribution of information and energy

Abstract Glasses-free three-dimensional (3D) display has attracted wide interest for providing stereoscopic virtual contents with depth cues. However, how to achieve high spatial and angular resolution while keeping ultrawide field of view (FOV) remains a significant challenge in 3D display. Here, we propose a light field 3D display with space-variant resolution for non-uniform distribution of information and energy. The spatial resolution of each view is modulated according to watching habit. A large-scale combination of pixelated 1D and 2D metagratings is used to manipulate dot and horizontal line views. With the joint modulation of pixel density and view arrangement, the information density and illuminance of high-demand views are at most 5.6 times and 16 times that of low-demand views, respectively. Furthermore, a full-color and video rate light field 3D display with non-uniform information distribution is demonstrated. The prototype provides 3D images with a high spatial resolution of 119.6 pixels per inch and a high angular resolution of 0.25 views per degree in the high-demand views. An ultrawide viewing angle of 140° is also provided. The proposed light field 3D display does not require ultrahigh-resolution display panels and has form factors of thin and light. Thus, it has the potential to be used in portable electronics, window display, exhibition display, as well as tabletop display.

Indentation-induced martensitic transformation in SMAs: Insights from phase-field simulations

Direct experimental characterization of indentation-induced martensitic microstructures in pseudoelastic shape memory alloys (SMAs) is not possible, and thus there is a lack of evidence and understanding regarding the microstructure pattern and related features. To fill this gap, in this work we employ the phase-field method to provide a detailed and systematic analysis of martensitic phase transformation during nanoindentation. A recently-developed finite-element-based computational model is used for this purpose, and a campaign of large-scale 3D simulations is carried out. First, the orientation-dependent indentation response in CuAlNi (a widely studied SMA) is examined. A detailed investigation of the predicted microstructures reveals several interesting features, some of them are consistent with theoretical predictions and some can be (to some extent) justified by experiments other than micro/nanoindentation. The results also highlight the key role of finite-deformation effects and elastic anisotropy of the phases on the model predictions. Next, a detailed study of indentation-induced martensitic transformation in NiTiPd (a potential low-hysteresis SMA) with varying Pd content is carried out. In terms of hysteresis, the results demonstrate the prevailing effect of the transformation volume change over phase compatibility in the conditions imposed by nanoindentation and emphasize on the dominant role of the interfacial energy at small scales. Results of such scope have not been reported so far.

CenterLoc3D: monocular 3D vehicle localization network for roadside surveillance cameras

Monocular 3D vehicle localization is an important task for vehicle behaviour analysis, traffic flow parameter estimation and autonomous driving in Intelligent Transportation System (ITS) and Cooperative Vehicle Infrastructure System (CVIS), which is usually achieved by monocular 3D vehicle detection. However, monocular cameras cannot obtain depth information directly due to the inherent imaging mechanism, resulting in more challenging monocular 3D tasks. Currently, most of the monocular 3D vehicle detection methods still rely on 2D detectors and additional geometric constraint modules to recover 3D vehicle information, which reduces the efficiency. At the same time, most of the research is based on datasets of onboard scenes, instead of roadside perspective, which is limited in large-scale 3D perception. Therefore, we focus on 3D vehicle detection without 2D detectors in roadside scenes. We propose a 3D vehicle localization network CenterLoc3D for roadside monocular cameras, which directly predicts centroid and eight vertexes in image space, and the dimension of 3D bounding boxes without 2D detectors. To improve the precision of 3D vehicle localization, we propose a multi-scale weighted-fusion module and a loss with spatial constraints embedded in CenterLoc3D. Firstly, the transformation matrix between 2D image space and 3D world space is solved by camera calibration. Secondly, vehicle type, centroid, eight vertexes, and the dimension of 3D vehicle bounding boxes are obtained by CenterLoc3D. Finally, centroid in 3D world space can be obtained by camera calibration and CenterLoc3D for 3D vehicle localization. To the best of our knowledge, this is the first application of 3D vehicle localization for roadside monocular cameras. Hence, we also propose a benchmark for this application including a dataset (SVLD-3D), an annotation tool (LabelImg-3D), and evaluation metrics. Through experimental validation, the proposed method achieves high accuracy with A{P_{3D}} of 51.30%, average 3D localization precision of 98%, average 3D dimension precision of 85% and real-time performance with FPS of 41.18.

3D Mapping for a Large Crane Using Rotating 2D-Lidar and IMU Attached to the Crane Boom

This paper describes a novel method for large-scale 3D mapping for construction cranes with an arbitrary motion of the sensor system (2D lidar and IMU) attached to the crane boom. A heavy lidar with a slowly rotating base is needed to make a large-scale map both vertically and horizontally for cranes. This sensor configuration and mapping conditions entail handling each 2D scan separately, making it difficult to adopt existing rotating 2D lidar-based methods that construct a virtual 3D scan from a set of 2D scans. In the proposed method, we introduce a complementary filter with moving average filtering for lidar pose estimation, which is more robust to severe vibration than Kalman filter-based methods. As there are only a small amount of overlaps between 2D lidar scans, we propose a map correction method based on a pose graph optimization with planar environmental constraints. We evaluate the proposed method in a simulation and a small-sized real environment and compare it with one of the state-of-the-art methods. The evaluation results reveal that the proposed method can accurately estimate the sensor poses, thereby generating a high-quality, large-scale 3D point cloud map.

Enhanced Combined Techniques for Interval-Valued Intuitionistic Fuzzy Multiple-Attribute Group Decision-Making and Applications to Quality Evaluation of Large-Scale Multi-View 3D Reconstruction

Enhanced Combined Techniques for Interval-Valued Intuitionistic Fuzzy Multiple-Attribute Group Decision-Making and Applications to Quality Evaluation of Large-Scale Multi-View 3D Reconstruction

A Category-Contrastive Guided-Graph Convolutional Network Approach for the Semantic Segmentation of Point Clouds

The semantic segmentation of light detection and ranging (LiDAR) point clouds plays an important role in 3D scene intelligent perception and semantic modeling. The unstructured, sparse and uneven characteristics of point clouds pose great challenges to the representation of the local geometric shapes, which degrades semantic segmentation performance. To address the challenges of describing local geometric shapes due to unstructured and sparse 3D point clouds, this paper proposes a category-contrastive-guided graph convolutional network (CGGC-Net) for the semantic segmentation of LiDAR point clouds. First, a detailed geometric structure of the raw point clouds is encoded to represent the inherent geometric pattern within the local neighborhood. At the same time, the geometric structures information is transmitted across multiple layers, so that the geometric structure encoding information containing different receptive fields and richer neighborhood spatial structure can be aggregated. Following this, the graph convolution neural network uses the edge convolution layer to adaptively describe the semantic correlation between the query point and its neighboring points, and combines the attention mechanism to gather the surrounding feature information to the query point. As a result, the graph convolution neural network and attention mechanism are iteratively stacked for the aggregation and fusion of spatial context semantic information, to generate highly discriminative semantic feature representation. Finally, the super parameters of the model are learned through a multitask optimization strategy guided by category-aware contrastive loss and cross-entropy loss. Experiments are conducted on the public SemanticKITTI dataset and the Stanford Large-Scale 3D Indoor Spaces dataset to demonstrate the effectiveness and reliability of the proposed CGGC-Net from both quantitative and qualitative perspectives. The results indicate its capability of automatically classifying LiDAR point clouds, with a mean intersection-over-union (mIoU) of 58.4%. Moreover, multiple comparative experiments also demonstrate the superior performance of the proposed method, exceeding state-of-the-art methods.

Multi-Object Navigation using Potential Target Position Policy Function.

Visual object navigation is an essential task of embodied AI, which follows the user's demands to let the agent navigate to the goal objects. Previous methods often focus on single object navigation. However, in real life, human demands are generally continuous and multiple, requiring the agent to implement multiple tasks in sequence. These demands can be addressed by repeatedly performing previous single task methods. However, dividing multiple tasks into several independent tasks to perform, without the global optimization between different tasks, the agents' trajectories may overlap, reducing the efficiency of navigation. In this paper, we propose an efficient reinforcement learning framework with a hybrid policy for multi-object navigation, aiming to maximally eliminate the noneffective actions. First, the visual observations are embedded to detect the semantic entities (such as objects). And the detected objects are memorized and projected into semantic maps, which can also be regarded as a long-term memory of the observed environment. Then a hybrid policy consisting of exploration and long-term planning strategies is proposed to predict the potential target position. In particular, when the target is directly oriented, the policy function makes long-term planning to the target based on the semantic map, which is implemented by a sequence of motion actions. In the alternative, when the target is not oriented, the policy function estimates an object potential position towards exploring the most possible objects (positions) that have close relations to the target. The relation between different objects is obtained with prior knowledge, which is used to predict the potential target position by integrating with the memorized semantic map. And then a path to the potential target is planned by the policy function. We evaluate our proposed method on two large-scale 3D realistic environment datasets, Gibson and Matterport3D, and the experimental results demonstrate the effectiveness and generalization of the proposed method.