检索结果-内蒙古大学图书馆

SKD: Keypoint Detection for Point Clouds Using Saliency Estimation

IEEE ROBOTICS AND AUTOMATION LETTERS 2021年第2期6卷 3785-3792页

作者： Tinchev, Georgi Penate-Sanchez, Adrian Fallon, Maurice Univ Oxford Oxford Robot Inst Cambridge CB2 9HQ England

We present SKD, a novel keypoint detector that uses saliency to determine the best candidates from a point cloud for tasks such as registration and reconstruction. The approach can be applied to any differentiable deep learning descriptor by using the gradients of that descriptor with respect to the 3D position of the input points as a measure of their saliency. The saliency is combined with the original descriptor and context information in a neural network, which is trained to learn robust keypoint candidates. The key intuition behind this approach is that keypoints are not extracted solely as a result of the geometry surrounding a point, but also take into account the descriptor's response. The approach was evaluated on two large LIDAR datasets - the Oxford RobotCar dataset and the KITTI dataset, where we obtain up to 50% improvement over the state-of-the-art in both matchability and repeatability. When performing sparse matching with the keypoints computed by our method we achieve a higher inlier ratio and faster convergence.

关键词： deep learning for visual perception recognition visual learning

来源：评论

学校读者我要写书评

暂无评论

Line as a visual Sentence: Context-Aware Line Descriptor for visual Localization

引用

IEEE ROBOTICS AND AUTOMATION LETTERS 2021年第4期6卷 8726-8733页

作者： Yoon, Sungho Kim, Ayoung Korea Adv Inst Sci & Technol Robot Program Daejeon South Korea SNU Dept Mech Engn Seoul South Korea

Along with feature points for image matching, line features provide additional constraints to solve visual geometric problems in robotics and computer vision (CV). Although recent convolutional neural network (CNN)-based line descriptors are promising for viewpoint changes or dynamic environments, we claim that the CNN architecture has innate disadvantages to abstract variable line length into the fixed-dimensional descriptor. In this letter, we effectively introduce Line-Transformers dealing with variable lines. Inspired by natural language processing (NLP) tasks where sentences can be understood and abstracted well in neural nets, we view a line segment as a sentence that contains points (words). By attending to well-describable points on a line dynamically, our descriptor performs excellently on variable line length. We also propose line signature networks sharing the line's geometric attributes to neighborhoods. Performing as group descriptors, the networks enhance line descriptors by understanding lines' relative geometries. Finally, we present the proposed line descriptor and matching in a Point and Line Localization (PL-Loc). We show that the visual localization with feature points can be improved using our line features. We validate the proposed method for homography estimation and visual localization.

关键词： Transformers visualization Image segmentation Simultaneous localization and mapping Location awareness Task analysis Convolutional neural networks Localization SLAM deep learning for visual perception

来源：评论

学校读者我要写书评

暂无评论

DA4Event: Towards Bridging the Sim-to-Real Gap for Event Cameras Using Domain Adaptation

引用

IEEE ROBOTICS AND AUTOMATION LETTERS 2021年第4期6卷 6616-6623页

作者： Planamente, Mirco Plizzari, Chiara Cannici, Marco Ciccone, Marco Strada, Francesco Bottino, Andrea Matteucci, Matteo Caputo, Barbara Politecn Torino Ist Italiano Tecnol I-16163 Genoa Italy Politecn Torino Dipartimento Automat & Informat DAUIN I-10129 Turin Italy Politecn Milan Dipartimento Elettron Informaz & Bioingn DEIB I-20133 Milan Italy

Event cameras are novel bio-inspired sensors, which asynchronously capture pixel-level intensity changes in the form of "events". The innovative way they acquire data presents several advantages over standard devices, especially in poor lighting and high-speed motion conditions. However, the novelty of these sensors results in the lack of a large amount of training data capable of fully unlocking their potential. The most common approach implemented by researchers to address this issue is to leverage simulated event data. Yet, this approach comes with an open research question: how well simulated data generalize to real data? To answer this, we propose to exploit, in the event-based context, recent Domain Adaptation (DA) advances in traditional computer vision, showing that DA techniques applied to event data help reduce the sim-to-real gap. To this purpose, we propose a novel architecture, which we call Multi-View DA4E (MV-DA4E), that better exploits the peculiarities of frame-based event representations while also promoting domain invariant characteristics in features. Through extensive experiments, we prove the effectiveness of DA methods and MV-DA4E on N-Caltech101. Moreover, we validate their soundness in a real-world scenario through a cross-domain analysis on the popular RGB-D Object Dataset (ROD), which we extended to the event modality (RGB-E).

关键词： visual learning deep learning for visual perception

来源：评论

学校读者我要写书评

暂无评论

Temporal Siamese Networks for Clutter Mitigation Applied to Vision-Based Quadcopter Formation Control

引用

IEEE ROBOTICS AND AUTOMATION LETTERS 2021年第1期6卷 32-39页

作者： Dunn, James Tron, Roberto Boston Univ Dept Elect & Comp Engn 8 St Marys St Boston MA 02215 USA Dept Mech & Syst Engn 8 St Marys St Boston MA 02215 USA

Convolutional neural networks applied to video streams often suffer from short-lived misclassifications or false alarms from clutter and noise. We introduce a novel network training method based on the Siamese Networks technique that mitigates false alarms in an Hourglass CNN that segments out quadcopters in a live video stream. To demonstrate this method in a real-world application in real-time, we implement it as part of a quadcopter tracker for vision-based formation control. Quadcopter drone formation control is an important capability for fields like area surveillance, search and rescue, agriculture, and reconnaissance. Of particular interest is formation control in environments where wireless communications and/or GPS may be either denied or not sufficiently accurate for the desired application. Using vision to guide the quadcopters addresses these situations, but computer vision algorithms are often computationally expensive and suffer from high false detection rates. Our novel Siamese networks-based clutter mitigation technique is a good way to mitigate this clutter without added computational complexity at run-time. We run our real-time implementation on an ODROID XU4 with a standard webcam mounted to a quadcopter drone. Flight tests in a motion capture volume demonstrate successful formation control with two quadcopters in a leader-follower setup.

关键词： Clutter Image segmentation Drones Training visualization Feeds Cameras Aerial systems deep learning for visual perception perception and autonomy visual tracking

来源：评论

学校读者我要写书评

暂无评论

Scanline Resolution-Invariant Depth Completion Using a Single Image and Sparse LiDAR Point Cloud

引用

IEEE ROBOTICS AND AUTOMATION LETTERS 2021年第4期6卷 6961-6968页

作者： Ryu, Kwonyoung Lee, Kang-il Cho, Jegyeong Yoon, Kuk-Jin Pohang Univ Sci & Technol Grad Sch Artificial Intelligence Comp Vis Lab Gyeongsangbuk Do 37673 South Korea Korea Adv Inst Sci & Technol Visual Intelligence Lab Mech Engn Daejeon 34141 South Korea Korea Adv Inst Sci & Technol Visual Intelligence Lab Mech Engn Daejeon 34141 South Korea

Most existing deep learning-based depth completion methods are only suitable for high (e.g. 64-scanline) resolution LiDAR measurements, and they usually fail to predict a reliable dense depth map with low resolution (4, 8, or 16-scanline) LiDAR. However, it is of great interest to reduce the number of LiDAR channels in many aspects (cost, weight of a device, power consumption). In this letter, we propose a new depth completion framework with various LiDAR scanline resolutions, which performs as well as methods built for 64-scanline resolution LiDAR inputs. For this, we define a consistency loss between the predictions from LiDAR measurements of different scanline resolutions. (i.e., 4, 8, 16, 32-scanline LiDAR measurements) Also, we design a fusion module to integrate features from different modalities. Experiments show our proposed method outperforms the current state-of-the-art depth completion methods for input LiDAR measurements of low scanline resolution and performs comparably to the methods(models) for input LiDAR measurements of 64-scanline resolution on the KITTI benchmark dataset.

关键词： Depth completion depth estimation sensor fusion deep learning for visual perception LiDAR

来源：评论

学校读者我要写书评

暂无评论

Elevation Angle Estimation in 2D Acoustic Images Using Pseudo Front View

引用

IEEE ROBOTICS AND AUTOMATION LETTERS 2021年第2期6卷 1535-1542页

作者： Wang, Yusheng Ji, Yonghoon Liu, Dingyu Tsuchiya, Hiroshi Yamashita, Atsushi Asama, Hajime Univ Tokyo Dept Precis Engn Grad Sch Engn Tokyo 1138654 Japan Japan Adv Inst Sci & Technol Grad Sch Adv Sci & Technol Nomi Ishikawa 9231211 Japan Wakachiku Construct Co Ltd Res Inst Chiba 2990268 Japan

A novel method to estimate the missing dimension in 2D acoustic images for 3D reconstruction is proposed in this paper. Acoustic cameras can acquire high resolution 2D images in underwater environment insusceptible to water turbidity and light condition. However, the formulation of acoustic images leads to the missing dimension problem. Estimating the unknown elevation angle dimension is a difficult task which has recently drawn the attention of researchers. The non-bijective characteristic between 3D points and 2D pixels increases the complexity of the problem. In this paper, a novel elevation angle estimation method is proposed. The method transfers the acoustic view to pseudo front view using a deep neural network. The proposed network can estimate the missing dimension and resolve the non-bijection problem of the 2D-3D correspondence. Because of the difficulty of acquiring depth information in underwater environments, the network is trained using simulated images. To mitigate the sim-real gap, a neural style transfer method is implemented to generate a realistic image dataset for training. Simulation experiments were carried out for evaluation and real data proved the feasibility of the proposed method.

关键词： Marine robotics deep learning for visual perception deep learning methods

来源：评论

学校读者我要写书评

暂无评论

S²Net: Accurate Panorama Depth Estimation on Surface

引用

IEEE ROBOTICS AND AUTOMATION LETTERS 2023年第2期8卷 1053-1060页

作者： Li, Meng Wang, Senbo Yuan, Weihao Shen, Weichao Sheng, Zhe Dong, Zilong Alibaba Grp Hangzhou 310000 Peoples R China

Monocular depth estimation is an ambiguous problem, thus global structural cues play an important role in current data-driven single-view depth estimation methods. Panorama images capture the complete spatial information of their surroundings utilizing the equirectangular projection which introduces large distortion. This requires the depth estimation method to be able to handle the distortion and extract global context information from the image. In this paper, we propose an end-to-end deep network for monocular panorama depth estimation on a unit spherical surface. Specifically, we project the feature maps extracted from equirectangular images onto unit spherical surface sampled by uniformly distributed grids, where the decoder network can aggregate the information from the distortion-reduced feature maps. Meanwhile, we propose a global cross-attention-based fusion module to fuse the feature maps from skip connection and enhance the ability to obtain global context. Experiments are conducted on five panorama depth estimation datasets, and the results demonstrate that the proposed method substantially outperforms previous state-of-the-art methods. All related codes will be open-sourced in the upcoming days.

关键词： Estimation Distortion Feature extraction Decoding Convolution Transformers Image representation deep learning for visual perception omnidirectional vision deep learning methods

来源：评论

学校读者我要写书评

暂无评论

PTP: Parallelized Tracking and Prediction With Graph Neural Networks and Diversity Sampling

引用

IEEE ROBOTICS AND AUTOMATION LETTERS 2021年第3期6卷 4640-4647页

作者： Weng, Xinshuo Yuan, Ye Kitani, Kris Carnegie Mellon Univ Robot Inst Pittsburgh PA 15206 USA

Multi-object tracking (MOT) and trajectory prediction are two critical components in modern 3D perception systems that require accurate modeling of multi-agent interaction. We hypothesize that it is beneficial to unify both tasks under one framework in order to learn a shared feature representation of agent interaction. Furthermore, instead of performing tracking and prediction sequentially which can propagate errors from tracking to prediction, we propose a parallelized framework to mitigate the issue. Also, our parallel track-forecast framework incorporates two additional novel computational units. First, we use a feature interaction technique by introducing Graph Neural Networks (GNNs) to capture the way in which agents interact with one another. The GNN is able to improve discriminative feature learning for MOT association and provide socially-aware contexts for trajectory prediction. Second, we use a diversity sampling function to improve the quality and diversity of our forecasted trajectories. The learned sampling function is trained to efficiently extract a variety of outcomes from a generative trajectory distribution and helps avoid the problem of generating duplicate trajectory samples. We evaluate on KITTI and nuScenes datasets showing that our method with socially-aware feature learning and diversity sampling achieves new state-of-the-art performance on 3D MOT and trajectory prediction. Project website is: http://***/projects/PTP.

关键词： Trajectory Feature extraction Forecasting Three-dimensional displays Predictive models Solid modeling Cultural differences Computer vision for transportation deep learning for visual perception visual tracking

来源：评论

学校读者我要写书评

暂无评论

Diversity-Aware Label Distribution learning for Microscopy Auto Focusing

引用

IEEE ROBOTICS AND AUTOMATION LETTERS 2021年第2期6卷 1942-1949页

作者： Zhang, Chuyan Gu, Yun Yang, Jie Yang, Guang-Zhong Shanghai Jiao Tong Univ Inst Med Robot Shanghai Peoples R China Shanghai Jiao Tong Univ Inst Image Proc & Pattern Recognit Shanghai Peoples R China Shanghai Ctr Brain Sci & Brain Inspired Technol Shanghai Peoples R China

Optical microscopy imaging is the gold standard for the diagnosis of cancers since it allows the cell-level visualization of tissues. The high quality of imaging is largely determined by the focus distances between the lens and objects. Therefore, a robust and efficient auto focusing algorithm is required to obtain the optimal focus position, especially for the robot-assisted microscopy systems. In this letter, we propose a diversity-aware learning framework to predict the optimal focus position based on a single image, without any reference. For robust and accurate estimation, the two-point representation of distance to the optimal focus position is utilized for label distribution learning. To reduce the intra-class variation caused by the diversity of pathological slides, we present a intra-class discrepancy penalty term following the composite-loss and the gradient-domain input strategy to concentrate more on image focus quality. Experiments on real microscopy datasets demonstrate that the proposed method achieves the promising performance in terms of accuracy, real-time and generalization. The mean absolute error is 0.308 $\mu$m, which is within the depth-of-field of the microscope. It outperforms the previous no-reference approaches by 39%.

关键词： Computer vision for medical robotics deep learning for visual perception

来源：评论

学校读者我要写书评

暂无评论

Self-Supervised learning of Domain-Invariant Local Features for Robust visual Localization Under Challenging Conditions

引用

IEEE ROBOTICS AND AUTOMATION LETTERS 2021年第2期6卷 2753-2760页

作者： Venator, Moritz El Himer, Yassine Aklanoglu, Selcuk Bruns, Erich Maier, Andreas Volkswagen Car SW Org D-85057 Ingolstadt Germany Friedrich Alexander Univ Erlangen Nurnberg Pattern Recognit Lab D-91058 Erlangen Germany Tech Univ Munich D-80333 Munich Germany

visual localization provides the basis for many robotics applications such as autonomous navigation or augmented reality. Especially in outdoor scenes, robust localization requires local features which can be reliably extracted and matched under changing conditions. Previous approaches have applied generative image-to-image translation models to align images in a single domain before correspondence search. In this letter, we invert the concept and elaborate why it is more promising to use image domain adaptation for training of robust local features. Integrating this idea into a self-supervised training framework, we show in various experiments covering image matching, visual localization, and scene reconstruction that our Domain-Invariant SuperPoint (DISP) outperforms existing self-supervised methods in terms of repeatability, generalization, and robustness. In contrast to competitive supervised local features, our modular and fully self-supervised approach can be easily adapted to different domains and localization tasks as it does not require ground truth correspondences for training.

关键词： Computer vision for transportation deep learning for visual perception localization

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：