Object Geometry Research Articles

A Novel Approach for Dynamic (4d) Multi-View Stereo System Camera Network Design.

Image network design is a critical factor in image-based 3D shape reconstruction and data processing (especially in the application of combined SfM/MVS methods). This paper aims to present a new approach to designing and planning multi-view imaging networks for dynamic 3D scene reconstruction without preliminary information about object geometry or location. The only constraints are the size of defined measurement volume, the required resolution, and the accuracy of geometric reconstruction. The proposed automatic camera network design method is based on the Monte Carlo algorithm and a set of prediction functions (considering accuracy, density, and completeness of shape reconstruction). This is used to determine the camera positions and orientations and makes it possible to achieve the required completeness of shape, accuracy, and resolution of the final 3D reconstruction. To assess the accuracy and efficiency of the proposed method, tests were carried out on synthetic and real data. For a set of 20 virtual images of rendered spheres, completeness of shape reconstruction was up by 92.3% while maintaining accuracy and resolution at the user-specified level. In the case of the real data, the differences between predictions and evaluations for average density were in the range between 33.8% to 45.0%.

An interactive sketch-based CAD interface realizing geometrical and topological editing across multiple objects based on fuzzy logic

With the widespread use of digital pen tablets and touch displays, sketch-based interactive computer-aided design (CAD) user interfaces have attracted considerable attention. In an existing interface called the Sketch Input Tracer (SKIT), however, the editing operation performed via an over-sketching stroke can target only a single geometric object at any given time. To overcome this limitation, in this study, we propose a new general-purpose sketch-based interface developed for use in two-dimensional CAD systems. In the proposed interface, a sketch-based editing operation can simultaneously modify the geometry and topology of multiple geometric objects via over-sketching. The interface was developed based on the fuzzy logic-based strategy of SKIT. With this interface, a user can draw in a creative manner; for example, they can begin with a rough sketch and progressively create a detailed design through repeated over-sketches. The results of experiments conducted demonstrate that experts can control the proposed interactive drawing interface and beginners can habituate to controlling it.

Continuous control actions learning and adaptation for robotic manipulation through reinforcement learning

This paper presents a learning-based method that uses simulation data to learn an object manipulation task using two model-free reinforcement learning (RL) algorithms. The learning performance is compared across on-policy and off-policy algorithms: Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC). In order to accelerate the learning process, the fine-tuning procedure is proposed that demonstrates the continuous adaptation of on-policy RL to new environments, allowing the learned policy to adapt and execute the (partially) modified task. A dense reward function is designed for the task to enable an efficient learning of the agent. A grasping task involving a Franka Emika Panda manipulator is considered as the reference task to be learned. The learned control policy is demonstrated to be generalizable across multiple object geometries and initial robot/parts configurations. The approach is finally tested on a real Franka Emika Panda robot, showing the possibility to transfer the learned behavior from simulation. Experimental results show 100% of successful grasping tasks, making the proposed approach applicable to real applications.

Identifying illumination conditions most suitable for attitude detection in light curves of simple geometries

• Illumination conditions determine the detectability of attitude state. • Alignment between illumination and observation vectors is poor for determination. • Offsets between illumination and observation vectors increases distinguishability. • Ideal illumination conditions depend on the position of rotation vector. • Light curve attitude determination benefits from a range of illumination conditions. The objective of this paper is to identify the illumination conditions that maximise the differences that can be measured between light curves of an object resulting from its attitude state. This is relevant to attitude determination techniques using light curve data, and is valuable for the design of observation strategies that maximise the information contained in light curves. For this purpose, synthetic light curves were generated for a number of attitude states, object geometries and illumination configurations. The light curves were generated through application of a Bidirectional Reflectance Distribution Function (BRDF) to a faceted object geometry. The differences between the light curves were quantified using a Root Mean Square Error (RMSE). Results showed that, depending on the reflection model, the object geometry and the attitude state, particular illumination conditions existed that led to the largest RMSE between different attitude states. In most cases, increasing the phase angle increased the RMSE between light curves arising from different attitude states. The maximum RMSE occurred when the illumination vector was either aligned with the rotation vector or offset from it by 90 ° . It is concluded that characterising the rotational motion of an object from its brightness data is best performed using multiple observations. These observations should be constructed in a way that maximises the difference in the illumination geometry. One way of achieving this would be to use observations from multiple observatories with a diverse range of geographical locations.

Regularizing deep networks with label geometry for accurate object localization on small training datasets

Localization is a critical subtask in object detection, which is closely related to spatial information of objects. Most current detectors simply rely on the fitting ability of deep neural networks to regress towards numerical targets such as coordinates of object boxes. Training deep networks for sufficient fitting ability requires a large number of annotations that are expensive to obtain. In this work, we fully exploit limited annotations by extracting label geometry to improve localization performance on small datasets. We generate distance transform of bounding box edges according to localization labels, with which we supervise intermediate outputs of networks pixel by pixel to reconstruct object geometry for localization. Distance transform is sensitive to box edges and provides geometric gradients flowing into boundaries. We learn such gradients to enhance geometric-aware features through a coupled training with regression, and use it to refine regressed boxes in an evolutionary manner in inference. Extensive experiments are implemented to demonstrate the effectiveness of our method. Our method can be applied in applications that required human-machine interaction, such as the driver-assistance system in autonomous driving, by providing accurate detections to assist humans in making better decisions.

Tomographic Volumetric Additive Manufacturing of Silicon Oxycarbide Ceramics

Ceramics are highly chemically, thermally, and mechanically resistant. These remarkable propertiers make them useful across multiple industries; but also, difficult to mold into complex shapes. A possibility to make convoluted ceramic parts is to use preceramic polymers (PCPs) in liquid form. The PCP resin is first solidified in a desired geometry and then transformed into ceramic compounds through a pyrolysis step that preserves the shape. Light‐based additive manufacturing (AM) is a promising route to achieve solidification of the PCP resin. Different approaches, such as stereolithography, have already been proposed but they all rely on a layer‐by‐layer printing process which limits printing speed and object geometry. Herein, the fabrication of complex 3D centimeter‐scale ceramic parts by using tomographic volumetric printing is presented, which is fast and offers a high‐resolution and geometrical design freedom. First, a photosensitive polysiloxane preceramic resin that is solidified by projecting light patterns from multiple angles is formulated. Then, the obtained 3D printed parts are converted into ceramics by pyrolysis. The strength of this approach is demonstrated through the fabrication of smooth, dense microcomponents exhibiting overhangs and hollow geometries without the need of supporting structures. Their resistance to thermal stress and harsh chemical treatments is characterized.

Improving Road Surface Area Extraction via Semantic Segmentation with Conditional Generative Learning for Deep Inpainting Operations

The road surface area extraction task is generally carried out via semantic segmentation over remotely-sensed imagery. However, this supervised learning task is often costly as it requires remote sensing images labelled at the pixel level, and the results are not always satisfactory (presence of discontinuities, overlooked connection points, or isolated road segments). On the other hand, unsupervised learning does not require labelled data and can be employed for post-processing the geometries of geospatial objects extracted via semantic segmentation. In this work, we implement a conditional Generative Adversarial Network to reconstruct road geometries via deep inpainting procedures on a new dataset containing unlabelled road samples from challenging areas present in official cartographic support from Spain. The goal is to improve the initial road representations obtained with semantic segmentation models via generative learning. The performance of the model was evaluated on unseen data by conducting a metrical comparison where a maximum Intersection over Union (IoU) score improvement of 1.3% was observed when compared to the initial semantic segmentation result. Next, we evaluated the appropriateness of applying unsupervised generative learning using a qualitative perceptual validation to identify the strengths and weaknesses of the proposed method in very complex scenarios and gain a better intuition of the model’s behaviour when performing large-scale post-processing with generative learning and deep inpainting procedures and observed important improvements in the generated data.

Microwave Coincidence Imaging Based on Attributed Scattering Model

In this letter, a novel microwave coincidence imaging (MCI) method based on the attributed scattering model (ASM) is proposed. Unlike the classical MCI model which assumes the target as a set of discrete scatterers, the ASM-based MCI equation contains three kinds of reference matrices corresponding to the point-scatterers (PSs), the line-segment-scatterers (LSSs) and the rectangular-plate-scatterers (RPSs), respectively. Hence the ASM-based MCI could resolve richer information about the object geometries. By solving the imaging equation via the alternating direction method of multipliers (ADMM) algorithm, the scattering coefficients will be obtained and the target can be reconstructed according to the presetting parameter sets. Meantime, the ASM-based MCI also earns the superresolution ability like the classical MCI, which is brought in by the temporal-spatial orthogonal radiation field. Simulations and experiment are carried out to demonstrate the performance and superresolution ability of proposed method. The ASM-based MCI makes contributions to the progress of radar forward-looking imaging theory and technology.

Shape-Adaptive Lighting for Complex Geometries in Automated Visual Inspection

Automating visual inspection has the potential to remove the current bottleneck of burdensome and subjective manual inspection in precision manufacturing. However, complex geometries of parts hinder uniform illumination, and high reflectivity challenges accurate focusing for digital visual data collection. This paper provides a novel method to distribute consistent illuminance for quality image data collection over complex-shaped, highly reflective surfaces. The central approach entails using arrays of independently controlled light sources to generate different lighting patterns, structures, and colors. Such results consider the geometry and 3D pose of objects in the environment and the surface topography of the work-piece to be inspected, leading to greater imaging qualities for precision inspection. This paper discusses the mathematical problem formulation, the analytic solution, the optimality of the proposed shape-adaptive lighting, and experimental results in imaging curved parts common in precision manufacturing.

Виды предгеометрий кубических теорий

The description of the types of geometries is one of the main problems in the structural classification of algebraic systems. In addition to the well-known classical geometries, a deep study of the main types of pregeometries and geometries was carried out for classes of strongly minimal and 𝜔-stable structures. These studies include, first of all, the works of B.I. Zilber and G. Cherlin, L. Harrington, A. Lachlan in the 1980s. Early 1980s B.I. Zilber formulated the well-known conjecture that the pregeometries of strongly minimal theories are exhausted by the cases of trivial, affine, and projective pregeometries. This hypothesis was refuted by E. Hrushovski, who proposed an original construction of a strongly minimal structure that is not locally modular and for which it is impossible to interpret an infinite group. The study of types of pregeometries continues to attract the attention of specialists in modern model theory, including the description of the geometries of various natural objects, in particular, Vamos matroids. In this paper we consider pregeometries for cubic theories with algebraic closure operator. And we notice that for pregeometries ⟨𝑆, acl⟩ in cubic theories, the substitution property holds if and only if the models of the theory do not contain infinite cubes, in particular, when there are no finite cubes of unlimited dimension. By virtue of this remark, we introduce new concepts of 𝑐-dimension, 𝑐-pregeometry, 𝑐-triviality, 𝑐- modularity, 𝑐-projectivity and 𝑐-locally finiteness. And besides, we prove the dichotomy theorem for the types of 𝑐-pregeometries.

THE SET OF INCOMPLETE SUMS OF THE MODIFIED GUTHRIE-NYMANN SERIES

In this paper we study topological and metric properties of the set of incomplete sums for positive series $\sum {a_k}$, where $a_{2n-1}=3/4^n+3/4^{in}$ and $a_{2n}=2/4^n+2/4^{in}$, $n \in N$. The series depends on positive integer parameter $i \geq 2$ and it is some perturbation of the known Guthrie-Nymann series. We prove that the set of incomplete sums of this series is a Cantorval (which is a specific union of a perfect nowhere dense set of zero Lebesgue measure and an infinite union of intervals), and its Lebesgue measure is given by formula: $\lambda(X^+_i)=1+\frac{1}{4^i-3}.$ The main idea of ??proving the theorem is based on the well-known Kakey theorem, the closedness of sets of incomplete sums of the series and the density of the set everywhere in a certain segment. The work provides a full justification of the facts for the case $i=2$. To justify the main facts, the ratio between the members and the remainders of the series is used. For $i=2$ we have $r_0=\sum {a_k}=2$, $a_{2n}-r_{2n}= \frac{1}{3} \cdot \frac{1}{4^n} + \frac{5}{3} \cdot \frac{1}{16^n}$ $r_{2n-1}-a_{2n-1}= \frac{2}{3} \cdot \frac{ 1}{4^n}-\frac{2}{3} \cdot \frac{1}{16^n}$. The relevance of the study of the object is dictated by the problems of the geometry of numerical series, fractal analysis and fractal geometry of one-dimensional objects and the theory of infinite Bernoulli convolutions, one of the problems of which is the problem of the singularity of the convolution of two singular distributions.

Electromagnetic Finite-Element Modeling of Induction Effects for Buried Objects in Magnetic Soils

A frequency-based finite element (FE) framework has been developed to predict and understand the response of an electromagnetic induction (EMI) sensor due to buried targets. The EMI sensor is used to detect buried targets in magnetic nonconducting soils. The framework was verified with an analytical model that utilizes dipole approximations. The framework was then used to predict the electromagnetic (EM) response due to interrelated stimuli and properties. The results indicate that the sensor was not sensitive to small variations (0–200 mm) in the standoff height and lateral positions, and only showed a significant change in the response due to stand-off variations that were greater than 200 mm. This low sensitivity to minor variations in standoff height and lateral position signify that there are critical distances related to the EM response of buried objects. The response to different target conductivities and permeabilities was also investigated for steel and aluminum targets. The lower conductivity steel targets had EM responses, where the inductive limit was reached at higher frequencies than the higher conductivity aluminum targets. Variations in target permeabilities for steel showed that as permeabilities increased, the frequencies at which the inductive limit was reached also increased. This verified predictive approach can provide a methodology to characterize the EM response of buried objects for a broad class of buried object EM properties, geometries, and input stimuli.

A Pansharpening Based on the Non-Subsampled Contourlet Transform and Convolutional Autoencoder: Application to QuickBird Imagery

This paper presents a pansharpening technique based on the non-subsampled contourlet transform (NSCT) and convolutional autoencoder (CAE). NSCT is exceptionally proficient at presenting orientation information and capturing the internal geometry of objects. First, it’s used to decompose the multispectral (MS) and panchromatic (PAN) images into high-frequency and low-frequency components using the same number of decomposition levels. Second, a CAE network is trained to generate original low-frequency PAN images from their spatially degraded versions. Low-resolution multispectral images are then fed into the trained convolutional autoencoder network to generate estimated high-resolution multispectral images. Third, another CAE network is trained to generate original high-frequency PAN images from their spatially degraded versions. The result of low-pass CAE is fed to the trained high-pass CAE to generate estimated high-resolution multispectral images. The final pan-sharpened image is accomplished by injecting the detailed map of the spectral bands into the corresponding estimated high-resolution multispectral bands. The proposed method is tested on QuickBird datasets and compared with some existing pan-sharpening techniques. Objective and subjective results demonstrate the efficiency of the proposed method.

Моделирование нестационарной теплопередачи в многослойной биметаллической пластине

When using the method of active thermal non-destructive testing to detect latent defects in multilayer bimetallic products, the adequate choice of technological modes of testing is of great importance. One of the important parameters is the optimum observation time, at which a sufficient amplitude of the temperature signal over the defect is provided. This value depends on the controlled object geometry, its thermophysical characteristics, and also selected testing modes and procedures. Various experimental stu-dies are used to determine the optimum observation time. Preparing bimetallic plates samples for full-scale experiments is a technologically complex and expensive task. Therefore, a large amount of preparatory research is carried out in the form of a numerical experiment, during which, with certain assumptions, changing the thermal state of the control object is simulated. Aim. Determination of optimal parameters of the thermal NDT process for a three-layer steel-aluminum plate with a planar delamination defect between the metal layers. For different variants of mutual location of the temperature registration device and source of heat flux of different power it is necessary to determine the heating time at which the sufficient amplitude of temperature signal is observed. Materials and methods. The methods of mathematical and computer modeling were used. Solving the task by the finite element method with the Agros2D code is given. Results. The calculation data have been obtained, which are used as the basis for constructing the temperature signal distribution over the bimetallic plate surface above the defect under different heating modes, heat flux rate, relative position of the heat stimulation source and temperature registration device, as well as determining the temperature signal dependence over the defect on the heating time. Conclusion. On the basis of modeling the temperature dependences necessary for interpretation of the results of thermal NDT of multilayered bimetallic plates have been determined, as well as the relations between the parameters of the heating process, which contribute to obtaining a temperature signal of sufficient amplitude. The results can be used for further experimental studies of the active thermal NDT processes of multilayer bimetallic materials.

High Degree of Freedom Aerodynamic Shape Optimization with Design-Space Dimensionality Reduction

In aerodynamic multi-objective optimization for object geometry, design variables are often multidimensional and weakly independent, which reduces the efficiency of optimization. In this study, a design space reduction method using ANOVA and AIC was proposed and implemented, then evaluated by applying it to a multi-objective aerodynamic optimal design problem. The results show that the proposed design space reduction method can extract important variables appropriately and enables fast, efficient, and accurate solution search.

INVESTIGATION OF POINTNET FOR SEMANTIC SEGMENTATION OF LARGE-SCALE OUTDOOR POINT CLOUDS

Abstract. Semantic segmentation of point clouds is indispensable for 3D scene understanding. Point clouds have credibility for capturing geometry of objects including shape, size, and orientation. Deep learning (DL) has been recognized as the most successful approach for image semantic segmentation. Applied to point clouds, performance of the many DL algorithms degrades, because point clouds are often sparse and have irregular data format. As a result, point clouds are regularly first transformed into voxel grids or image collections. PointNet was the first promising algorithm that feeds point clouds directly into the DL architecture. Although PointNet achieved remarkable performance on indoor point clouds, its performance has not been extensively studied in large-scale outdoor point clouds. So far, we know, no study on large-scale aerial point clouds investigates the sensitivity of the hyper-parameters used in the PointNet. This paper evaluates PointNet’s performance for semantic segmentation through three large-scale Airborne Laser Scanning (ALS) point clouds of urban environments. Reported results show that PointNet has potential in large-scale outdoor scene semantic segmentation. A remarkable limitation of PointNet is that it does not consider local structure induced by the metric space made by its local neighbors. Experiments exhibit PointNet is expressively sensitive to the hyper-parameters like batch-size, block partition and the number of points in a block. For an ALS dataset, we get significant difference between overall accuracies of 67.5% and 72.8%, for the block sizes of 5m × 5m and 10m × 10m, respectively. Results also discover that the performance of PointNet depends on the selection of input vectors.

Per-pixel displacement mapping using cone tracing with correct silhouette

Per-pixel displacement mapping is a texture mapping technique that adds the microrelief effect to 3D surfaces without increasing the density of their corresponding meshes. This technique relies on ray tracing algorithms to find the intersection point between the viewing ray and the microrelief stored in a 2D texture called a depth map. This intersection makes it possible to deter-mine the corresponding pixel to produce an illusion of surface displacement instead of a real one. Cone tracing is one of the per-pixel displacement map-ping techniques for real-time rendering that relies on the encoding of the empty space around each pixel of the depth map. During the preprocessing stage, this space is encoded in the form of top-opened cones and then stored in a 2D texture, and during the rendering stage, it is used to converge more quickly to the intersection point. Cone tracing technique produces satisfactory results in the case of flat surfaces, but when it comes to curved surfaces, it does not support the silhouette at the edges of the 3D mesh, that is to say, the relief merges with the surface of the object, and in this case, it will not be rendered correctly. To overcome this limitation, we have presented two new cone tracing algorithms that allow taking into consideration the curvature of the 3D surface to determine the fragments belonging to the silhouette. These two algorithms are based on a quadratic approximation of the object geometry at each vertex of the 3D mesh. The main objective of this paper is to achieve a texture mapping with a realistic appearance and at a low cost so that the rendered objects will have real and complex details that are visible on their entire surface and without modifying their geometry. Based on the ray-tracing algorithm, our contribution can be useful for current graphics card generation, since the programmable units and the frameworks associated with the new graphics cards integrate today the technology of ray tracing

Characterizing and structuring urban GIS data for housing stock energy modelling and retrofitting

Addressing the gap between available urban building data, energy performance evaluation, and spatial distribution of end-users is a growing concern of municipalities to support energy planning and urban retrofitting towards low carbon emission. These days, many existing urban building energy models (UBEM) use the non-geo-specified 3D building stock geometries or the digital twins of cities as their primary input for energy simulation. The digital twins of cities or the built environment is an accurate scale virtual representation (geo-specified geometry and information) of city objects enabled to be paired with dynamic predictive models of their energy performance. The use of built environment digital twins for their energy assessment works well for those cities whose digital twin datasets are readily available for public or private use. In the cities without digital twin datasets, their energy assessment is limited by tools that use 2D building stock footprints as their primary input or apply a top-down urban energy simulation approach. This gap motivated the current research to identify an extended UBEM data preparation workflow focused on generating city digital twins using accessible GIS datasets as well as national and local data sources. The workflow is majorly developed to deal with GIS data processing utilizing multilevel spatial data integration and refinement to fill in the recognized inconsistencies of building databases. Then, procedural 3D city modelling and an automated transformation to a semantic, CityGML format are successfully provided to generate digital twin of the study area in an open data model. The archetype modelling method is also employed to create a set of customized housing energy profiles to support tailored bottom-up UBEM and retrofitting scenarios relying on the simulation platform Insel4Cities, under development at Concordia University. The workflow was practically carried out on a case study, Kelowna city in British Columbia (BC)/Canada, to validate the district heating load. The average simulated heating energy use intensity showed a deviation of less than 2.5% from the BC measured data. As an example of workflow usability, the developed model was also utilized to assess a retrofit scenario, combining a decentralized heat pump and photovoltaic system, to present the district's potential for net-zero action planning.

Image-based attitude determination of co-orbiting satellites using deep learning technologies

Active debris removal missions pose demanding guidance, navigation and control requirements. We present a novel approach which adopts deep learning technologies to the problem of attitude determination of an uncooperative debris satellite of an a-priori unknown geometry. A siamese convolutional neural network is developed, which detects and tracks inherently useful landmarks from sensor data, after training upon synthetic datasets of visual, LiDAR or RGB-D data. The method is capable of real-time performance while improving upon conventional computer vision-based approaches, and generalises well to previously unseen object geometries, enabling this approach to be a feasible solution for safely performing guidance and navigation in active debris removal, satellite servicing and other close proximity operations. The performance of the algorithm, its sensitivity to model parameters and its robustness to illumination and shadowing conditions, are analysed via numerical simulation.

Low-latency time-of-flight non-line-of-sight imaging at 5 frames per second

Non-Line-Of-Sight (NLOS) imaging aims at recovering the 3D geometry of objects that are hidden from the direct line of sight. One major challenge with this technique is the weak available multibounce signal limiting scene size, capture speed, and reconstruction quality. To overcome this obstacle, we introduce a multipixel time-of-flight non-line-of-sight imaging method combining specifically designed Single Photon Avalanche Diode (SPAD) array detectors with a fast reconstruction algorithm that captures and reconstructs live low-latency videos of non-line-of-sight scenes with natural non-retroreflective objects. We develop a model of the signal-to-noise-ratio of non-line-of-sight imaging and use it to devise a method that reconstructs the scene such that signal-to-noise-ratio, motion blur, angular resolution, and depth resolution are all independent of scene depth suggesting that reconstruction of very large scenes may be possible.