检索结果-内蒙古大学图书馆

Distributed Bundle Adjustment Based on Penalty Function Method

IEEE Robotics and Automation Letters 2025年第6期10卷 5449-5456页

作者： Sun, Qixuan Zhang, Guanghui Zhu, Dongchen Wang, Lei Zhang, Xiaolin Li, Jiamao Chinese Academy of Sciences Bionic Vision System Laboratory State Key Laboratory of Transducer Technology Shanghai Institute of Microsystem and Information Technology Shanghai200050 China University of Chinese Academy of Sciences Beijing100049 China

Bundle Adjustment (BA) aims to estimate the camera poses and build maps utilizing the nonlinear optimization algorithm. The update step of the optimization is obtained by solving a linear system, which is the bottleneck of the BA efficiency. Many works perform bundle adjustment in a distributed manner to reduce the computational cost of solving the linear system and achieve outstanding performance. However, the distributed methods implement the constrained optimization method that involves the computation of the Lagrange multiplier, causing extra overhead. To improve the efficiency of BA, we propose a novel distributed algorithm based on the penalty method, dubbed PenBA, that formulates an unconstrained problem and hence saves the computation of extra variables. To solve the unconstrained linear system with low cost, we design a novel distributed Preconditioned Conjugate Gradient (PCG) which has complexity bounded by hyper-parameters. It can be proved that, given the parameter update strategy of the optimization method, the penalty linear system can be solved with fast convergence. Experimental results in the BAL dataset illustrate that our penalty method outperforms the Lagrange multiplier method and conventional BA. Compared with the most accurate conventional PCG-based BA solver, PenBA decreases the objective values by 3.8%, the time per iteration by 17% and the maximum memory consumption by 24%. Experimental results in the SfM and SLAM datasets prove that PenBA applies to both computer vision and robotics contexts. © 2016 IEEE.

关键词： Constrained optimization

来源：评论

学校读者我要写书评

暂无评论

Improved calibration and 3D reconstruction for micro fringe projection profilometry

引用

Optics Express 2025年第6期33卷 13455-13471页

作者： Zhang, Chenfan Zhu, Dongchen Shi, Wenjun Wang, Lei Li, Jiamao Bionic Vision System Laboratory State Key Laboratory of Transducer Technology Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Shanghai200050 China University of Chinese Academy of Sciences Beijing100049 China

Micro fringe projection profilometry (MFPP) is a crucial three-dimensional measurement technology widely used in industrial inspection. As the system measurement accuracy is limited by calibration and reconstruction errors, this paper proposes a joint calibration of telecentric camera and projector, and develops noise-robust 3D reconstruction, ultimately forming a high-precision MFPP for microscopic object measurement. Specifically, the optimal parameters for the telecentric camera model are determined by constructing a virtual 3D calibration target, which circumvents the limitations that plane-based calibration methods often produce local optimal solutions. In response to the presence of phase errors, a method employing global homography is proposed to refine the projector pixel coordinates of feature points, thus enhancing the calibration accuracy of the projector. Furthermore, we integrate epipolar constraints into the reconstruction process, which helps eliminate noise points and optimizes the reconstruction results by constructing epipolar deviation. Extensive experiments demonstrate that our MFPP system can achieve micron-level measurement accuracy. © 2025 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement.

关键词： Profilometry

来源：评论

学校读者我要写书评

暂无评论

MemoFlow: Modifying Explicit Motion of Inconsistency in Optical Flow 27th

MemoFlow: Modifying Explicit Motion of Inconsistency in Op...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Wang, Mengfei Shi, Wenjun Zhu, Dongchen Wang, Lei Li, Jiamao Bionic Vision System Laboratory State Key Laboratory of Transducer Technology Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Shanghai200050 China University of Chinese Academy of Sciences Beijing100049 China

ISBN: (纸本)9783031781124

Due to the limitation of training datasets and global motion understanding to optical flow estimation, current methods only focus on local clues and ignore the motion continuity of consecutive frames, resulting in an inconsistent motion problem. After theoretical analysis, we find the multi-frame methods need to pay more attention to the continuity between cost volumes than two-frame methods. Thus, our method is based on a multi-frame framework and introduces extra object coordinates in each frame by the segmentation model to revise the matching pairs in cost volume. Specifically, we introduce a Cost Volume Adaptation Module, including a Bbox Spatial Queries to store coordinates information and a Correlation Query Queue to query the object position of different frames. On the Sintel and KITTI test benchmark, our proposed MemoFlow achieves 1.00 and 1.69 average endpoint error (AEPE) on the clean and final passes and an F1-all error of 4.43%, ranking 1st among all three-frame methods and two-frame methods. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Optical correlation

来源：评论

学校读者我要写书评

暂无评论

CRFENet: Cross-Spectral Stereo Matching Network Exploiting Reciprocal Feature Evolution 2nd

CRFENet: Cross-Spectral Stereo Matching Network Exploiting R...

引用

2nd International Conference on the Frontiers of Robotics and Software Engineering, FRSE 2024

作者： Li, Yangqin Zhang, Guanghui Zhu, Dongchen Shi, Wenjun Wang, Lei Zhang, Xiaolin Shanghai Institute of Microsystem and Information Technology Bionic Vision System Laboratory State Key Laboratory of Transducer Technology Chinese Academy of Sciences Shanghai200050 China University of Chinese Academy of Sciences Beijing100049 China School of Information Science and Technology ShanghaiTech University Shanghai201210 China

ISBN: (纸本)9789819630127

Cross-spectral stereo matching opens up a new route for disparity estimation in low-light scenes. However, the variations in light absorption and reflection properties for objects at different wavelengths create substantial domain gaps in cross-spectral image pairs. Accurately matching pixel pairs with significant domain gaps remains challenging. In this work, we propose a novel lightweight cross-spectral stereo matching network, dubbed CRFENet, exploiting reciprocal feature evolution. It enables the complementary evolution of the disparity prediction network and spectral translation network. To facilitate feature interactions between different domains to improve disparity results in regions with significant domain gaps, we design a domain mutual rail (DMR) module, which integrates monocular spectral and binocular cross-spectral features. Furthermore, we present a domain gap consistency constraint and corresponding confidence-weighted loss function, which aim to alleviate the matching challenge in regions with significant domain gaps through decoupling domain gap and disparity. Experimental results indicate that our CRFENet achieves state-of-the-art performance on the PittsStereo RGB-NIR and DrivingStereo datasets, especially for challenging regions with significant domain gaps. Besides, the model parameters are only 3.6M, which means high practicality. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Stereocenters

来源：评论

学校读者我要写书评

暂无评论

Optimized Design and Calibration of a Human-Eye-Sized Active Binocular vision system Based on Spherical Parallel Mechanism

引用

IEEE Robotics and Automation Letters 2025年

作者： Wang, Kaifang Yang, DongDong Zhang, Li Liu, Jun Zhang, Xiaolin Chinese Academy of Sciences Bionic Vision System Laboratory State Key Laboratory of Transducer Technology Shanghai Institute of Microsystem and Information Technology Shanghai200050 China University of Chinese Academy of Sciences Beijing100049 China Anhui Eyevolution Technology Co. Ltd. Shanghai200050 China ShanghaiTech University Shanghai201210 China

The Active Binocular vision system (ABVS), resembling the human eye, demonstrates potential for improving visual perception in robotic systems, especially in dynamic and complex environments. In this letter, we present an optimized design of a three degree-of-freedom (DoF) Active Monocular vision system (AMVS) based on a Spherical Parallel Manipulator (SPM). By combining two identical AMVS units, we form an ABVS, which has been successfully integrated into a humanoid robotic head. Due to the highly nonlinear kinematics of SPM and complex error coupling in its multi-link structure, traditional end-to-end neural network training methods are insufficient in accuracy and require large datasets. To address these challenges, we propose a two-branch optimization network that significantly improves calibration accuracy. Furthermore, we introduce a four-branch fine-tuning strategy that enables accurate kinematic models to be obtained with only a small amount of data from new AMVS devices. Experimental results demonstrate that the two-branch optimization network reduces rotational prediction error by 16% and translational error by 5% compared to a single-branch network. Furthermore, the four-branch fine-tuning network achieves comparable accuracy to a fully trained single-branch network using only 343 data points. Finally, our ABVS shows the capability to perform 3D visual tasks, such as stereo reconstruction during movement. © 2016 IEEE.

关键词： Anthropomorphic robots

来源：评论

学校读者我要写书评

暂无评论

Points and Patches in Textures by Exploring in Monocular Depth Estimation

Points and Patches in Textures by Exploring in Monocular Dep...

引用

Computer, vision and Intelligent Technology (ICCVIT), International Conference on

作者： Deming Wu Guanghui Zhang Wei Ming Wenhao Wang Xiaochen Jing Dongchen Zhu Bionic Vision System Laboratory Shanghai Institude of Microsystem and Information Technology Shanghai China

ISBN: (数字)9798331540043

ISBN: (纸本)9798331540050

Rich texture information is sometimes not good for deep learning, and can introduce unwanted noise information, especially for monocular depth learning. In order to solve the phenomenon of ‘messy textures’ caused by rich texture in monocular depth estimation, this paper exploits the property of intra-patch disparity consistency to reconstruct the image to mitigate this phenomenon. The joint training method using patch matching and single-point matching is verified on Cityscapes and KITTI Engie Split datasets to achieve better results.

关键词： Training Deep learning Computer vision Depth measurement Semantics Noise Kernel Image reconstruction

来源：评论

学校读者我要写书评

暂无评论

Visual-LiDAR Fusion Relocation for SLAM systems

Visual-LiDAR Fusion Relocation for SLAM Systems

引用

2023 International Conference on Frontiers of Robotics and Software Engineering, FRSE 2023

作者： Liu, Zexi Liu, Yanqing Wang, Xin Zhu, Dongchen Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Bionic Vision System Laboratory State Key Laboratory of Transducer Technology Shanghai200050 China University of Chinese Academy of Sciences Beijing100049 China Wuxi Institute of Inspection Testing and Certification Wuxi214000 China

ISBN: (纸本)9798350301113

Relocation is a fundamental part of the mobile robot positioning system, and the robot can obtain the initial position through relocation. However, commonly used visual relocation schemes are susceptible to lighting, weather, and seasonal changes. Therefore, this paper proposes a relocation method for vision and LiDAR sensor fusion. By making full use of different sensors, this method can also achieve good relocation performance in scenarios where the individual sensor is degraded. Our algorithm consists of three parts, namely visual relocation, LiDAR relocation and point cloud registration verification. The Bag-of-Words method is used for visual relocation, the Scan Context method is used for LiDAR relocation, and the ICP or NDT method is used for point cloud registration verification. It is worth noting that each part of our algorithm is an independent module, so it can be easily replaced by other methods, which greatly improves the flexibility of the algorithm. We tested our algorithm on KITTI, M2DGR, and NCLT datasets, and the experimental results show that visual-LiDAR fusion relocation has greatly improved performance compared to visual relocation. When environmental factors such as light, weather, and seasons change, visual-LiDAR fusion relocation can still operate robustly, which is the key to long-term robot positioning. © 2023 IEEE.

关键词： Optical radar

来源：评论

学校读者我要写书评

暂无评论

Discretized Channel Access Mechanism Based on CSMA Contention and TDMA Reservation

Discretized Channel Access Mechanism Based on CSMA Contentio...

引用

Information Communication Technologies Conference (ICTC)

作者： Ruijun Shu Ligang Liu Jiang Wang Xinyue Zhang Bionic Vision System Laboratory Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Shanghai China

ISBN: (数字)9781728167763

ISBN: (纸本)9781728167770

With the rapid growth of multimedia services in ad-hoc networks, how to transmit a mass of data efficiently becomes an important problem. Aiming to achieve more efficient channel utilization in ad-hoc networks, this paper presents a channel resource allocation scheme by combining CSMA/CA (Carrier Sense Multiple Access with Collision Avoidance) with TDMA (Time Division Multiple Access). In the proposed scheme, nodes are allowed to randomly access the channel and reserve timeslots for multiple times, and the access process only starts at the beginning of a timeslot. Moreover, the reservation length of each node can be adjusted adaptively according to the network load and requirements of the node. Simulation results show that the proposed approach has better performance in terms of the collision probability and throughput in distributed networks, especially in the scenarios with a heavy load and a large number of nodes.

关键词： Time division multiple access Ad hoc networks Resource management Data communication Simulation Throughput

来源：评论

学校读者我要写书评

暂无评论

A Queue-Length and Collision-Risk-Prediction Based Scheduling for Wireless Mesh Networks

A Queue-Length and Collision-Risk-Prediction Based Schedulin...

引用

Information Communication Technologies Conference (ICTC)

作者： Ruijun Shu Huiyue Yi Ligang Liu Dan Liu Bionic Vision System Laboratory Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Shanghai China

ISBN: (数字)9781728167763

ISBN: (纸本)9781728167770

The hidden terminals lead to frequent conflicts in the Wireless Mesh Network (WMNs), which degrades the system performance. Moreover, as the number of network nodes increases, the conflict becomes worse. In order to improve the throughput, this paper analyzes conflict issues in distributed WMNs and proposes a scheduling scheme based on queue-length and collision risk. In this scheme, the node exchanges the queue-length information in real-time, and the node that has the longest queue-length in the two-hop range can send data. Moreover, the node changes the sending order of the packets with the same priority by predicting the collision risk of the packet before sending, and preferentially sends the packet with the smallest collision risk. Finally, simulation results show that the proposed scheme can effectively increase the throughput when the network load is heavy, and it can significantly decrease the dropped packets and the delay.

关键词： Throughput Topology Delays Media Access Protocol Wireless communication Multiaccess communication Network topology

来源：评论

学校读者我要写书评

暂无评论

Improved Token-Ring-Based MAC Scheme for Two-Hop Ad Hoc Network

Improved Token-Ring-Based MAC Scheme for Two-Hop Ad Hoc Netw...

引用

International Conference on Communication Software and Networks, ICCSN

作者： Ruijun Shu Cheng Shen Bionic Vision System Laboratory Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Shanghai China

ISBN: (数字)9781728198156

ISBN: (纸本)9781728198163

In this paper, we investigate the time slots allocation scheme for two-hop networks. In order to decrease waste of time slots and reduce end-to-end delay, an improved scheme is proposed based on the token-based adaptive medium access control (TA-MAC) scheme. In the proposed scheme, we redesigned the superframe structure. Specifically, the scheme divides the superframe into four subframes and specifies the permissible transmission behavior of each subframe. There are five tokens circulated separately among nodes in each token ring for packet transmissions. For each token ring, when a node holds a token, it is assigned a slot for transmitting a packet, and then transfers the token to another node. Furthermore, with the objective function of minimizing end-to-end delay, an optimization model is established to obtain the optimal system parameters. Theoretical analysis and simulation results show that the proposed scheme maintains a high throughput rate, ensures the fairness of resource allocation, and reduces the end-to-end delay.

关键词：

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：