Nowadays, 3dreconstruction pipelines based on structure-from-motion and multi-view stereo techniques can reconstruct meticulous, large-scale geometry models and high-resolution textures. However, due to the shadow pa...
详细信息
ISBN:
(纸本)9781728147529
Nowadays, 3dreconstruction pipelines based on structure-from-motion and multi-view stereo techniques can reconstruct meticulous, large-scale geometry models and high-resolution textures. However, due to the shadow parts of the acquired images, the reconstructed texture map has dark areas, which is not convenient for relighting, which limits the use of the reconstruction model in VR/AR programs or games. In this paper, we introduce a large-scale synthetic city intrinsic images dataset for evaluating intrinsic decomposition of city scenes. Given our database, we developed a data-driven method for obtaining the diffuse albedo of reconstructed scenes, our method achieves good results in our intrinsic images dataset and real images.
In order to improve the efficiency and effect of outdoor 3dreconstruction, we designed this image set preprocessing system, which can screen out some pictures that do not meet the requirements and are of low quality ...
详细信息
ISBN:
(纸本)9781728147529
In order to improve the efficiency and effect of outdoor 3dreconstruction, we designed this image set preprocessing system, which can screen out some pictures that do not meet the requirements and are of low quality through screening the blur degree, exposure degree and jitter degree of UAV aerial photo set. Using filtered photo sets can greatly reduce the impact of low-quality images on modeling speed and quality.
This paper presents a novel approach to construct spatially-referenced, multidimensional virtual city models from remotely generated point clouds for areas that lack reliable geographical reference data. A multidimens...
详细信息
ISBN:
(纸本)9783030588113;9783030588106
This paper presents a novel approach to construct spatially-referenced, multidimensional virtual city models from remotely generated point clouds for areas that lack reliable geographical reference data. A multidimensional point cloud is an unstructured array of single, irregular points in a spatial 3d coordinate system plus time stamp. If geospatial reference points are available, a point cloud is geo-referenced. Geo-referenced point clouds contain a high-precision reference dataset. Point clouds can be utilised in a variety of applications. They are particularly suitable for the representation of surfaces, structures, terrain and objects. Point clouds are used here to generate a virtual3d city model representing the complex, granular cityscape of Jerusalem and its centre, the Old City. The generation of point clouds is based on two data acquisition methods: active data capture by laser scanning and passive data collection by photogrammetric methods. In our case, very high-resolution stereo imagery in visible light and near infrared bands have been systematically acquired an aerial flight campaign. The spatio-temporal data gathered necessitate further processing to extract the geographical reference and semantic features required in a specific resolution and scale. An insight is given into the processing of an unstructured point cloud to extract and classify the 3d urban fabric and reconstruct its objects. Eventually, customised, precise and up-to-date geographical datasets can be made available for a defined region at a defined resolution and scale.
In recent decades, computer vision and computer graphics especially 3dreconstruction have been a hot area of computer research. As the scenes reconstructed from multi-view images become larger and larger, a single ma...
详细信息
ISBN:
(纸本)9781728147529
In recent decades, computer vision and computer graphics especially 3dreconstruction have been a hot area of computer research. As the scenes reconstructed from multi-view images become larger and larger, a single machine can no longer meet the requirement of the dense point cloudreconstruction in large-scale scene *** this paper, a distributed method is proposed to reconstruct the dense point cloud for accurate multi-view reconstruction. First, the initial image set is constructed into a graph, and the graph is divided into several graphs using the improved graph cut algorithm, so the image set will be parted into several small image sets according to the result of the graph cut. Then, the generation and optimization of dense point cloud will be performed on different nodes of the cluster. Finally, the dense point clouds generated on different machines will be merged on the primary node to generate a dense point cloud of the entire scene. Experiments on public large data sets and our own large-scale aerial photography show that the distributed method is fast, robust, and suitable for various large scene areas.
In this paper, we rethink the problem of scene reconstruction from an embodied agent’s perspective: While the classic view focuses on the reconstruction accuracy, our new perspective emphasizes the underlying functio...
详细信息
In this paper, we rethink the problem of scene reconstruction from an embodied agent’s perspective: While the classic view focuses on the reconstruction accuracy, our new perspective emphasizes the underlying functions and constraints such that the reconstructed scenes provide actionable information for simulating interactions with agents. Here, we address this challenging problem by reconstructing an interactive scene using RGB-ddata stream, which captures (i) the semantics and geometry of objects and layouts by a 3d volumetric panoptic mapping module, and (ii) object affordance and contextual relations by reasoning over physical common sense among objects, organized by a graph-based scene representation. Crucially, this reconstructed scene replaces the object meshes in the dense panoptic map with part-based articulated CAd models for finer-grained robot interactions. In the experiments, we demonstrate that (i) our panoptic mapping module outperforms previous state-of-the-art methods, (ii) a high-performant physical reasoning procedure that matches, aligns, and replaces objects’ meshes with best-fitted CAd models, and (iii) reconstructed scenes are physically plausible and naturally afford actionable interactions; without any manual labeling, they are seamlessly imported to ROS-based simulators andvirtual environments for complex robot task executions.1
In recent years, the world of computer graphics has made tremendous progress. Several 3d model visualization techniques have emerged and have been introduced in the hardware's. The machines on which are realized t...
详细信息
ISBN:
(纸本)9781450362894
In recent years, the world of computer graphics has made tremendous progress. Several 3d model visualization techniques have emerged and have been introduced in the hardware's. The machines on which are realized the 3dvisualization have also evolved. We do not need expensive computers to see a world in 3d;a simple computer can do the trick. This evolution has also created a demand for visualizing models that are more complex and realistic. Historically, research has focused on the development of 3d information and acquisition techniques from scenes and objects. demand has grown more in the field of computer graphics, virtual reality and communication. Acquiring 3d information's from real objects in a scene requires intricate calibration procedures every time the system is used. In addition, the use of these acquisition systems requires expertise by their users. This creates a significant demand for flexibility in acquisitions. These procedures must be absent or limited to a minimum. Because of these different factors, many techniques have been developed in recent years. Many of them only need a simple camera and a computer to be able to acquire 3d images.
The impressive performance of deep convolutional neural networks in single-view 3dreconstruction suggests that these models perform non-trivial reasoning about the 3d structure of the output space. Recent work has ch...
详细信息
Simultaneous Localization And Mapping (SLAM) is an important technique used in robotics, computer vision, andvirtual/augmented reality. SLAM algorithms have moved past creating sparse maps to making dense 3d reconstr...
详细信息
ISBN:
(纸本)9781728116013
Simultaneous Localization And Mapping (SLAM) is an important technique used in robotics, computer vision, andvirtual/augmented reality. SLAM algorithms have moved past creating sparse maps to making dense 3dreconstruction of the environment. dense SLAM algorithms have high computational demands that require hardware acceleration to be done efficiently in real-time. FPGAs are an attractive compute platform for SLAM systems as they are low power and high performance. Unfortunately, dense SLAM algorithms are complex and FPGAs are notoriously difficult to program. In this work, we study the best techniques for accelerating 3dreconstruction on FPGA. We analyze a 3dreconstruction system, and implement modular FPGA designs for the main components of this application. We target both an FPGA SoC and a larger FPGA PCIe board, and perform a design space exploration (dSE) of our designs. We analyze the results of our dSE, characterize the design spaces to highlight important features, and we implement the best designs in an open-source and end-to-enddense SLAM system running on a FPGA SoC board. On the SoC board, using the FPGA increases the throughput of the whole application by a factor of two compared to the ARM processor, and individual algorithms are up to 38 times faster on the FPGA.
Invariance and equivariance to the rotation group have been widely discussed in the 3ddeep learning community for pointclouds. Yet most proposed methods either use complex mathematical tools that may limit their acce...
详细信息
ISBN:
(纸本)9781665428132
Invariance and equivariance to the rotation group have been widely discussed in the 3ddeep learning community for pointclouds. Yet most proposed methods either use complex mathematical tools that may limit their accessibility, or are tied to specific input data types and network architectures. In this paper, we introduce a general framework built on top of what we call Vector Neuron representations for creating SO (3) -equivariant neural networks for pointcloud processing. Extending neurons from 1d scalars to 3d vectors, our vector neurons enable a simple mapping of SO (3) actions to latent spaces thereby providing a framework for building equivariance in common neural operations – including linear layers, non-linearities, pooling, and normalizations. due to their simplicity, vector neurons are versatile and, as we demonstrate, can be incorporated into diverse network architecture backbones, allowing them to process geometry inputs in arbitrary poses. despite its simplicity, our method performs comparably well in accuracy and generalization with other more complex and specialized state-of-the-art methods on classification and segmentation tasks. We also show for the first time a rotation equivariant reconstruction network. Source code is available at https://***/FlyingGiraffe/vnn.
暂无评论