摘要 :
In this paper, we put forward a new method for surface reconstruction from image-based point clouds. In particular, we introduce a new visibility model for each line of sight to preserve scene details without decreasing the noise ...
展开
In this paper, we put forward a new method for surface reconstruction from image-based point clouds. In particular, we introduce a new visibility model for each line of sight to preserve scene details without decreasing the noise filtering ability. To make the proposed method suitable for point clouds with heavy noise, we introduce a new likelihood energy term to the total energy of the binary labeling problem of Delaunay tetrahedra, and we give its s-t graph implementation. Besides, we further improve the performance of the proposed method with the dense visibility technique, which helps to keep the object edge sharp. The experimental result shows that the proposed method rivalled the state-of-the-art methods in terms of accuracy and completeness, and performed better with reference to detail preservation.
收起
摘要 :
Vectorized reconstruction from indoor point cloud has attracted increasing attention in recent years due to its high regularity and low memory consumption. Compared with aerial mapping of outdoor urban environments, indoor point c...
展开
Vectorized reconstruction from indoor point cloud has attracted increasing attention in recent years due to its high regularity and low memory consumption. Compared with aerial mapping of outdoor urban environments, indoor point cloud generated by LiDAR scanning or image-based 3D reconstruction usually contain more clutter and missing areas, which greatly increase the difficulty of vectorized reconstruction. In this paper, we propose an effective multistep pipeline to reconstruct vectorized models from indoor point cloud without the Manhattan or Atlanta world assumptions. The core idea behind our method is the combination of a sequence of 2D segment or cell assembly problems that are defined as global optimizations while reducing the reconstruction complexity and enhancing the robustness to different scenes. The proposed method includes a semantic segmentation stage and a reconstruction stage. First, we segment the permanent structures of indoor scenes, including ceilings, floors, walls and cylinders, from the input data, and then, we reconstruct these structures in sequence. The floorplan is first generated by detecting wall planes and selecting optimal subsets of projected wall segments with Integer Linear Programming (ILP), followed by constructing a 2D arrangement and recovering the ceiling and floor structures by Markov Random Filed (MRF) labeling on the arrangement. Finally, the wall structures are modeled by lifting each edge of the arrangement to a proper height by means of another global optimization. Merging the respective results yields the final model. The experimental results show that the proposed method could obtain accurate and compact vectorized models on both precise LiDAR data and defect-laden MVS data compared with other state-of-the-art approaches.
收起
摘要 :
In this paper, we present a simple yet effective rotation averaging pipeline, termed Incremental Rotation Averaging (IRA), which is inspired by the well-developed incremental Structure from Motion (SfM) techniques. Unlike the trad...
展开
In this paper, we present a simple yet effective rotation averaging pipeline, termed Incremental Rotation Averaging (IRA), which is inspired by the well-developed incremental Structure from Motion (SfM) techniques. Unlike the traditional rotation averaging methods which estimate all the absolute rotations simultaneously and focus on designing either robust loss function or outlier filtering strategy, here the absolute rotations are estimated in an incremental way. Similar to the incremental SfM, our IRA is robust to relative rotation outliers and could achieve accurate rotation averaging results. In addition, we propose several key techniques, such as initial triplet and Next-Best-View selection, Weighted Local/Global Optimization, and Re-Rotation Averaging, to push the rotation averaging results one step further. Ablation studies and comparison experiments on the 1DSfM, Campus, and San Francisco datasets demonstrate the effectiveness of our IRA and its advantages over the state-of-the-art rotation averaging methods in accuracy and robustness.
收起
摘要 :
To produce a complete 3D reconstruction of a large-scale architectural scene, both ground and aerial images are usually captured. A common approach is to first reconstruct the models from different image sources separately, and al...
展开
To produce a complete 3D reconstruction of a large-scale architectural scene, both ground and aerial images are usually captured. A common approach is to first reconstruct the models from different image sources separately, and align thetn afterwards. Using this pipeline, this work proposes an accurate and efficient approach for ground-to-aerial model alignment in a coarse-to-fine manner. First, both the ground model and aerial model are transformed into the geo-referenced coordinate system using GPS meta-information for coarse alignment. Then, the coarsely aligned models are refined by a similarity transformation that is estimated based on 3D point correspondences between them, and the 3D point correspondences are determined in a 2D-image-matching manner by considering the rich textural and contextual information in the 2D images. Due to the dramatic differences in viewpoint and scale between ground and aerial images, which make matching them directly nearly impossible, we perform an intermediate view-synthesis step to mitigate the matching difficulty. To this end, the following three key issues are addressed: (a) selecting a suitable subset of aerial images to cover the ground model properly; (b) synthesizing images from the ground model under the viewpoints of the selected aerial images; and finally, (c) obtaining the 2D point matches between the synthesized images and the selected aerial images. The experimental results show that the proposed model alignment approach is quite effective and outperforms several state-of-the-art techniques in terms of both accuracy and efficiency. (C) 2017 Elsevier Ltd. All rights reserved.
收起
摘要 :
Currently global structure-from-motion (SfM) pipeline consists of four steps: estimating camera rotations first, then computing camera positions, triangulating tracks, and finally doing bundle adjustment. However, for large-scale ...
展开
Currently global structure-from-motion (SfM) pipeline consists of four steps: estimating camera rotations first, then computing camera positions, triangulating tracks, and finally doing bundle adjustment. However, for large-scale SfM problems, the tracks are usually too noisy and redundant for the bundle adjustment. Thus in this work, we propose a novel fast tracks selection method to improve both efficiency and robustness of the bundle adjustment. Firstly, three selection criteria: Compactness, Accurateness, and Connectedness, are introduced, where the first two are to calculate a selection priority for each track and the third is to guarantee the completeness of scene structure. Then, to satisfy these criteria, a more informative subset of tracks is selected by covering multiple spanning trees of epipolar geometry graph. Since tracks selection acts only an intermediate step in the whole SfM pipeline, it can be in principle embedded into any global SfM pipelines. To validate the effectiveness of our tracks selection module, we insert it into a state-of-the-art global SfM system and compare it with three other selection methods. Extensive experiments show that by embedding our tracks selection module, the new SfM system performs similarly or better than the original one in terms of reconstruction completeness and accuracy, but is much more efficient and scalable for large-scale scene reconstructions. Finally, our tracks selection module is further embedded into two other global SfM systems to demonstrated its versatility. (C) 2017 Elsevier Ltd. All rights reserved.
收起
摘要 :
Incremental Structure-from-Motion (SfM) techniques have exhibited superior practicability in many recent studies; however, efficiency and robustness remain key challenges for these techniques. In this work, we propose a new increm...
展开
Incremental Structure-from-Motion (SfM) techniques have exhibited superior practicability in many recent studies; however, efficiency and robustness remain key challenges for these techniques. In this work, we propose a new incremental SfM method that overcomes these problems in a united framework that contains two iteration loops. The inner loop is a track selection loop, where a well-conditioned subset of the feature tracks is iteratively selected to accelerate the time-consuming bundle adjustment. The outer loop is a camera registration loop, where the a priori camera rotations are estimated via rotation averaging on multiple orthogonal maximum spanning trees (OMSTs) of the view-graph and used as weak supervision for the registration. The calibrated camera poses that agree with the a priori camera rotations are preferentially registered, and after all the consistent cameras have been calibrated, the remaining cameras are incrementally registered. The results of extensive experiments demonstrate that our system can reconstruct both general and ambiguous image datasets, and our system outperforms many state-of-the-art SfM systems in terms of efficiency and robustness.
收起
摘要 :
Localization and reconstruction are two highly related research areas. Both of them have developed rapidly in recent years. Apparently, with the help of ground and aerial meta-data integration, the performance of both localization...
展开
Localization and reconstruction are two highly related research areas. Both of them have developed rapidly in recent years. Apparently, with the help of ground and aerial meta-data integration, the performance of both localization and reconstruction can go a step further. For localization, aerial meta-data provides a global reference, by which the ground query can achieve a cumulative error free absolute localization. As for reconstruction, a complete and detailed model can be reconstructed by integrating ground and aerial meta-data. Though with many advantages, the integration itself is non-trivial. It is difficult to obtain ground-to-aerial correspondences neither in 2D manner nor in 3D manner. That is because: (1) The differences between the ground and aerial images in viewpoint, scale, illumination, etc. are notable; (2) The discrepancies between the ground and aerial point clouds in terms of point density, accuracy, noise level, etc. are very large. To deal with these problems, lots of methods have been proposed recently. In this paper, the methods of integrating ground and aerial meta-data for localization and reconstruction are reviewed respectively. Though many intermediate results with high quality have been achieved, we hope that inspired by the reviewed methods in this paper, more thorough methods and impressive results would emerge. (C) 2018 Elsevier B.V. All rights reserved.
收起
摘要 :
A view-graph is vital for both the accuracy and robustness of structure-from-motion (SfM). Conventional matrix decomposition techniques treat all edges of view-graph equally; hence, many edge outliers are produced in matching pair...
展开
A view-graph is vital for both the accuracy and robustness of structure-from-motion (SfM). Conventional matrix decomposition techniques treat all edges of view-graph equally; hence, many edge outliers are produced in matching pairs with fewer feature matches. To address this problem, we propose an incremental framework for view-graph construction, where the robustness of matched pairs that have a larger number of feature matches is propagated to their connected images. Given pairwise feature matches, a verified maximum spanning tree (VMST) is first constructed; for each edge in the VMST, we perform a local reconstruction and register its visible cameras. Based on the local reconstruction, pairwise relative geometries are computed and some new epipolar edges are produced. In this way, these newly computed edges inherit the robustness and accuracy of VMST, and by embedding them into VMST, our view-graph is constructed. We feed our view-graph into a standard SfM pipeline and compare this newly formed system with many of state-of-the-art SfM methods. The experimental results demonstrate that our view graph provides a better foundation for conventional SfM systems, and enables them to reconstruct both general and ambiguous images. ? 2020 Elsevier Ltd. All rights reserved.
收起
摘要 :
Multilabel image annotation is one of the most important open problems in computer vision field. Unlike existing works that usually use conventional visual features to annotate images, features based on deep learning have shown po...
展开
Multilabel image annotation is one of the most important open problems in computer vision field. Unlike existing works that usually use conventional visual features to annotate images, features based on deep learning have shown potential to achieve outstanding performance. In this work, we propose a multimodal deep learning framework, which aims to optimally integrate multiple deep neural networks pretrained with convolutional neural networks. In particular, the proposed framework explores a unified two stage learning scheme that consists of (i) learning to fine-tune the parameters of deep neural network with respect to each individual modality, and (ii) learning to find the optimal combination of diverse modalities simultaneously in a coherent process. Experiments conducted on a variety of public datasets evaluate the performance of the proposed framework for multilabel image annotation, in which the encouraging results validate the effectiveness of the proposed algorithms. (C) 2015 Elsevier B.V. All rights reserved.
收起