检索结果-内蒙古大学图书馆

SSRNet: Scalable 3D Surface Reconstruction Network

学校读者我要写书评

暂无评论

SSRNet: Scalable 3D Surface Reconstruction Network

Conference on Computer Vision and Pattern Recognition (CVPR)

作者： Zhenxing Mi Yiming Luo Wenbing Tao National Key Laboratory of Science and Technology on Multi-spectral Information Processing School of Artifical Intelligence and Automation Huazhong University of Science and Technology China

ISBN: (数字)9781728171685

ISBN: (纸本)9781728171692

Existing learning-based surface reconstruction methods from point clouds are still facing challenges in terms of scalability and preservation of details on large-scale point clouds. In this paper, we propose the SSRNet, a novel scalable learning-based method for surface reconstruction. The proposed SSRNet constructs local geometry-aware features for octree vertices and designs a scalable reconstruction pipeline, which not only greatly enhances the predication accuracy of the relative position between the vertices and the implicit surface facilitating the surface reconstruction quality, but also allows dividing the point cloud and octree vertices and processing different parts in parallel for superior scalability on large-scale point clouds with millions of points. Moreover, SSRNet demonstrates outstanding generalization capability and only needs several surface data for training, much less than other learning-based reconstruction methods, which can effectively avoid overfitting. The trained model of SSRNet on one dataset can be directly used on other datasets with superior performance. Finally, the time consumption with SSRNet on a large-scale point cloud is acceptable and competitive. To our knowledge, the proposed SSRNet is the first to really bring a convincing solution to the scalability issue of the learning-based surface reconstruction methods, and is an important step to make learning-based methods competitive with respect to geometry processing methods on real-world and challenging data. Experiments show that our method achieves a breakthrough in scalability and quality compared with state-of-the-art learning-based methods.

关键词： Three-dimensional displays Surface reconstruction Octrees Reconstruction algorithms Surface treatment Scalability Shape

Cascade Network with Guided Loss and Hybrid Attention for Two-view Geometry

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Chen, Zhi Yang, Fan Tao, Wenbing National Key Laboratory of Science and Technology on Multi-spectral Information Processing School of Artifical Intelligence and Automation Huazhong University of Science and Technology China

In this paper, we are committed to designing a high-performance network for two-view geometry. We first propose a Guided Loss and theoretically establish the direct negative correlation between the loss and Fn-measure by dynamically adjusting the weights of positive and negative classes during training, so that the network is always trained towards the direction of increasing Fn-measure. By this way, the network can maintain the advantage of the cross-entropy loss while maximizing the Fn-measure. We then propose a hybrid attention block to extract feature, which integrates the bayesian attentive context normalization (BACN) and channel-wise attention (CA). BACN can mine the prior information to better exploit global context and CA can capture complex channel context to enhance the channel awareness of the network. Finally, based on our Guided Loss and hybrid attention block, a cascade network 2 is designed to gradually optimize the result for more superior performance. Experiments have shown that our network achieves the state-of-the-art performance on benchmark datasets. Copyright © 2020, The Authors. All rights reserved.

关键词： Benchmarking

Infrared and Visible Image Matching Algorithm Based on SIFT and LDB 1

学校读者我要写书评

暂无评论

14th Conference on Image and Graphics Technologies and Applications, IGTA 2019

作者： Zhang, Lirui Dai, Min Tian, Jinwen National Key Laboratory of Science and Technology on Multi-spectral Information Processing Technology School of Automation Huazhong University of Science and Technology Wuhan430074 China

ISBN: (数字)9789811399176

ISBN: (纸本)9789811399169

A new local feature extraction method (BSPL) is proposed and applied to heterogeneous image matching to solve the problem that the traditional SIFT features have poor matching performance in heterogeneous image matching. A number of improvements have been made to ensure that common features of heterogeneous images can be extracted efficiently. The gradient histogram-equalized image is used as the input matching image;The bilateral filtering is used to construct the scale space pyramid to replace the Gaussian filtering of the traditional SIFT, which can make the details such as the edges of the image better preserved;PCA-based LDB descriptor is used as feature expression to improve the robustness of feature expression. Experimental results show that the proposed local feature descriptor has rotation and scale invariance, and effectively improves the number of matching points, matching accuracy, matching precision and matching adaptability, which is an effective infrared and visible image matching method. © 2019, Springer Nature Singapore Pte Ltd.

关键词： Image matching

P2B: Point-to-Box Network for 3D Object Tracking in Point Clouds

学校读者我要写书评

暂无评论

P2B: Point-to-Box Network for 3D Object Tracking in Point Cl...

Conference on Computer Vision and Pattern Recognition (CVPR)

作者： Haozhe Qi Chen Feng Zhiguo Cao Feng Zhao Yang Xiao National Key Laboratory of Science and Technology on Multi-Spectral Information Processing School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan China

ISBN: (数字)9781728171685

ISBN: (纸本)9781728171692

Towards 3D object tracking in point clouds, a novel point-to-box network termed P2B is proposed in an end-to-end learning manner. Our main idea is to first localize potential target centers in 3D search area embedded with target information. Then point-driven 3D target proposal and verification are executed jointly. In this way, the time-consuming 3D exhaustive search can be avoided. Specifically, we first sample seeds from the point clouds in template and search area respectively. Then, we execute permutation-invariant feature augmentation to embed target clues from template into search area seeds and represent them with target-specific features. Consequently, the augmented search area seeds regress the potential target centers via Hough voting. The centers are further strengthened with seed-wise targetness scores. Finally, each center clusters its neighbors to leverage the ensemble power for joint 3D target proposal and verification. We apply PointNet++ as our backbone and experiments on KITTI tracking dataset demonstrate P2B's superiority (~10%'s improvement over state-of-the-art). Note that P2B can run with 40FPS on a single NVIDIA 1080Ti GPU. Our code and model are available at https://***/HaozheQi/P2B.

关键词： Three-dimensional displays Proposals Object tracking Target tracking Two dimensional displays Feeds Machine learning

A correction method for aero-optics thermal radiation effects based on gradient distribution and dark channel

学校读者我要写书评

暂无评论

Optoelectronics Letters 2019年第5期15卷 374-380页

作者： HONG Han-yu SHI Yu ZHANG Tian-xu LIU Zhao Hubei Key Laboratory of Optical Information and Pattern Recognition Hubei Research Centre of Video Image and High Definition Projection School of Electrical and Information Engineering Wuhan Institute of Technology Wuhan 430074 China National Key Laboratory of Science and Technology on Multi-Spectral Information Processing School of Automation Huazhong University of Science and Technology Wuhan 430074 China

In this paper, a new algorithm is proposed to remove the effects of aerodynamic optical thermal radiation from a single infrared image. In this method, the joint probability model of gradient distribution is introduced by studying the "global smoothing and local fluctuation" characteristics of the bias field. A prior L0 norm of dark channel is introduced to constrain the latent clear image. Finally, the split Bregman method is used to solve the optimization problem. The effectiveness of the proposed method is verified by a series of experiments, and the results are compared with the results of the existing methods for the correction of thermal radiation effects.

关键词： aero-optics thermal radiation effects gradient distribution dark channel

P2B: Point-to-box network for 3D object tracking in point clouds

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Qi, Haozhe Feng, Chen Cao, Zhiguo Zhao, Feng Xiao, Yang National Key Laboratory of Science and Technology on Multi-Spectral Information Processing School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan430074 China

Towards 3D object tracking in point clouds, a novel point-to-box network termed P2B is proposed in an end-to-end learning manner. Our main idea is to first localize potential target centers in 3D search area embedded with target information. Then point-driven 3D target proposal and verification are executed jointly. In this way, the time-consuming 3D exhaustive search can be avoided. Specifically, we first sample seeds from the point clouds in template and search area respectively. Then, we execute permutation-invariant feature augmentation to embed target clues from template into search area seeds and represent them with target-specific features. Consequently, the augmented search area seeds regress the potential target centers via Hough voting. The centers are further strengthened with seed-wise targetness scores. Finally, each center clusters its neighbors to leverage the ensemble power for joint 3D target proposal and verification. We apply PointNet++ as our backbone and experiments on KITTI tracking dataset demonstrate P2B's superiority (∼10%'s improvement over state-of-the-art). Note that P2B can run with 40FPS on a single NVIDIA 1080Ti GPU. Our code and model are available at https://***/HaozheQi/P2B. Copyright © 2020, The Authors. All rights reserved.

关键词： Machine learning

Refinements in motion and appearance for online multi-object tracking

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Huang, Piao Han, Shoudong Zhao, Jun Liu, Donghaisheng Wang, Hongwei Yu, En Kot, Alex Chi Chung National Key Laboratory of Science and Technology on Multi-spectral Information Processing School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan430074 China Nanyang Technological University

Modern multi-object tracking (MOT) system usually involves separated modules, such as motion model for location and appearance model for data association. However, the compatible problems within both motion and appearance models are always ignored. In this paper, a general architecture named as MIF is presented by seamlessly blending the Motion integration, three-dimensional(3D) Integral image and adaptive appearance feature Fusion. Since the uncertain pedestrian and camera motions are usually handled separately, the integrated motion model is designed using our defined intension of camera motion. Specifically, a 3D integral image based spatial blocking method is presented to efficiently cut useless connections between trajectories and candidates with spatial constraints. Then the appearance model and visibility prediction are jointly built. Considering scale, pose and visibility, the appearance features are adaptively fused to overcome the feature misalignment problem. Our MIF based tracker (MIFT) achieves the state-of-the-art accuracy with 60.1 MOTA on both MOT16&17 challenges. Copyright © 2020, The Authors. All rights reserved.

关键词： Visibility

From open set to closed set: Supervised spatial divide-and-conquer for object counting

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Xiong, Haipeng Lu, Hao Liu, Chengxin Liu, Liang Shen, Chunhua Cao, Zhiguo National Key Laboratory of Science and Technology on Multi-Spectral Information Processing School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan430074 China University of Adelaide SA5005 Australia

Visual counting, a task that aims to estimate the number of objects from an image/video, is an open-set problem by nature, i.e., the number of population can vary in [0;+1) in theory. However, collected data and labeled instances are limited in reality, which means that only a small closed set is observed. Existing methods typically model this task in a regression manner, while they are prone to suffer from an unseen scene with counts out of the scope of the closed set. In fact, counting has an interesting and exclusive property-spatially decomposable. A dense region can always be divided until sub-region counts are within the previously observed closed set. We therefore introduce the idea of spatial divide-and-conquer (S-DC) that transforms open-set counting into a closed-set problem. This idea is implemented by a novel Supervised Spatial Divide-and-Conquer Network (SS-DCNet). Thus, SS-DCNet can only learn from a closed set but generalize well to open-set scenarios via S-DC. SS-DCNet is also efficient. To avoid repeatedly computing sub-region convolutional features, S-DC is executed on the feature map instead of on the input image. We provide theoretical analyses as well as a controlled experiment on toy data, demonstrating why closed-set modeling makes sense. Extensive experiments show that SS-DCNet achieves the state-of-the-art performance on three crowd counting datasets (ShanghaiTech, UCF_CC_50 and UCF-QNRF), a vehicle counting dataset (TRANCOS) and a plant counting dataset (MTC), with a 7:7% relative improvement on the UCF-QNRF, 33:1% on the TRANCOS, and 26:4% on the MTC. SS-DCNet also reports the state-of-the-art cross-domain performance on crowd counting datasets. Particularly in the task from UCF-QNRF to ShanghaiTech Part_A, SS-DCNet even beats most existing models trained directly on the target domain. Code and models have been made available at: Https://***/SS-DCNet. Copyright © 2020, The Authors. All rights reserved.

关键词： Computation theory

Semantic object and plane slam for rgb-d cameras 2nd

学校读者我要写书评

暂无评论

Semantic object and plane slam for rgb-d cameras

2nd Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2019

作者： Zheng, Longyu Tao, Wenbing National Key Laboratory of Science and Technology on Multi-spectral Information Processing School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan430074 China Shenzhen Huazhong University of Science and Technology Research Institute Shenzhen518057 China

ISBN: (纸本)9783030317256

Simultaneous Localization And Mapping (SLAM) is a fundamental problem in mobile robotics as well as in virtual reality (VR) and augmented reality (AR). Traditional visual-SLAM systems, such as ORB-SLAM, deal with sparse features extracted from high gradient image regions. While being robust and stable to light changing, camera rotation and scale changing to some extent, they are easy to accumulate drifts and cannot provide high-level information of the environment. To solve this problem, we take advantage of state of the art object detectors and a robust ICP method to build a map made up of objects providing semantic information as well as extra constraints to original point-based SLAM system. What’s more, for indoor scenes where planar structures are common, we introduce plane landmarks into SLAM framework to reduce drift. Experiments show that the our method can build a semantically more meaningful map without reducing the original SLAM system’s performance remarkably. © Springer Nature Switzerland AG 2019.

关键词： Object detection