The bio-inspired emerging dynamic vision sensor (DVS), characterized by its exceptional high temporal resolution and immediate response, possesses an innate advantage in capturing rapidly changing scenes. Nevertheless...
详细信息
The bio-inspired emerging dynamic vision sensor (DVS), characterized by its exceptional high temporal resolution and immediate response, possesses an innate advantage in capturing rapidly changing scenes. Nevertheless, it is also susceptible to severe noise interference, especially in challenging conditions like low illumination and high exposure. Notably, the existing noise processing approaches tend to oversimplify data into 2-dimensional (2D) patterns, disregarding the sparse and irregular crucial event structure information that the DVS intrinsically provides via its asynchronous output. Aiming at these problems, we propose a residual graph neural network (RGNN) framework based on density spatial clustering for event denoising, called DBRGNN. Leveraging the temporal window rule, we extract non-overlapping event segments from the DVS event stream and adopt a density-based spatial clustering algorithm to obtain event groups with spatial correlations. To fully exploit the inherent sparsity and plentiful spatiotemporal information of the raw event stream, we transform each event group as compact graph representations via directed edges and feed them into a graph coding module composed of a series of graph convolutional and pooling layers to learn robust geometric features from event sequences. Importantly, our approach effectively reduces noise levels without compromising the spatial structure and temporal coherence of spike events. Compared with other baseline methods, our DBRGNN achieves competitive performance by quantitative and qualitative evaluations on publicly available datasets under varying lighting conditions and noise ratios.
The Multi-mOdality Cut and pAste (MoCa) method cuts data from other frames and pastes it onto the current training data frame to increase the number of training object samples. However, the samples used by MoCa are al...
详细信息
ISBN:
(纸本)9789819743988;9789819743995
The Multi-mOdality Cut and pAste (MoCa) method cuts data from other frames and pastes it onto the current training data frame to increase the number of training object samples. However, the samples used by MoCa are all derived from the original dataset, which limits its ability to enhance object diversity. Recently, diffusion models have achieved remarkable results in the field of image generation, where simple prompts can enable the model to create entirely different paintings. In this paper, we propose DiffMoCa, which leverages the powerful creative capabilities of diffusion models to redraw the images cut by MoCa, thereby increasing the diversity of the objects and enhancing the generalization ability of the model. DiffMoCa demonstrates its capabilities in extensive experiments, wherein it surpasses MoCa by 2.2% in mAP on the KITTI dataset under moderate conditions.
Inspired by convolutional neural networks, graph convolutional networks (GCNs) have been proposed for processing non-Euclidean graph data and successfully been applied in recommendation systems, smart traffic, etc. Ho...
详细信息
Inspired by convolutional neural networks, graph convolutional networks (GCNs) have been proposed for processing non-Euclidean graph data and successfully been applied in recommendation systems, smart traffic, etc. However, subject to the sparsity and irregularity of GCN models, the complex execution pattern of large-scale GCN poses huge challenges to the efficient inference on general purpose CPUs and GPUs, such as workload imbalance and irregular memory access. Therefore, we propose a softwarehardware co-design framework for low-latency GCN inference on field programmable gate array. Specifically, at the algorithm level, we propose an attention-mechanism-based graph sparsification approach to reduce the redundant relation in the graph structure and alleviate irregularity without losing accuracy. Then, at the hardware design level, based on the sparsified graph, we propose a two-stage hardware architecture that supports the two phases with a distinct execution mode in the GCN. In order to achieve low-latency computation, edge-level and feature-level parallelism are exploited in the aggregation phase. In addition, a graph partition strategy is exploited to efficiently improve data reuse. The experimental results demonstrate that our proposed framework can achieve 739x speedup compared to CPU, 13.7x speedup compared to GPU on average and 6.8x speedup compared to state-of-the-art accelerators.
Natural untrimmed videos provide rich visual content for self-supervised learning. Yet most previous efforts to learn spatio-temporal representations rely on manually trimmed videos, such as Kinetics dataset (Carreira...
详细信息
Natural untrimmed videos provide rich visual content for self-supervised learning. Yet most previous efforts to learn spatio-temporal representations rely on manually trimmed videos, such as Kinetics dataset (Carreira and Zisserman 2017), resulting in limited diversity in visual patterns and limited performance gains. In this work, we aim to improve video representations by leveraging the rich information in natural untrimmed videos. For this purpose, we propose learning a hierarchy of temporal consistencies in videos, i.e., visual consistency and topical consistency, corresponding respectively to clip pairs that tend to be visually similar when separated by a short time span, and clip pairs that share similar topics when separated by a long time span. Specifically, we present a Hierarchical Consistency (HiCo++) learning framework, in which the visually consistent pairs are encouraged to share the same feature representations by contrastive learning, while topically consistent pairs are coupled through a topical classifier that distinguishes whether they are topic-related, i.e., from the same untrimmed video. Additionally, we impose a gradual sampling algorithm for the proposed hierarchical consistency learning, and demonstrate its theoretical superiority. Empirically, we show that HiCo++ can not only generate stronger representations on untrimmed videos, but also improve the representation quality when applied to trimmed videos. This contrasts with standard contrastive learning, which fails to learn powerful representations from untrimmed videos. Source code will be made available here.
This article is devoted to addressing the cluster formation-containment tracking problem of networked robotic systems (NRSs) with unknown model uncertainties and disturbances under directed graphs. A novel hierarchica...
详细信息
This article is devoted to addressing the cluster formation-containment tracking problem of networked robotic systems (NRSs) with unknown model uncertainties and disturbances under directed graphs. A novel hierarchical prescribed-time extended state observer (ESO) based control algorithm is developed such that all the robotic systems are divided into multiple subgroups. For any subgroup, the master nodes form the different desired formation shapes at specific time points and the center of the shape follows the trajectory of the related target. Moreover, the follower nodes converge into the corresponding formation shapes. In the estimator loop, a cluster time-varying formation-containment tracking (TVFCT) algorithm is designed by employing a time-varying function such that the cluster formation shape can be guaranteed. In the local control loop, an extended state observer is employed to estimate the total disturbances (model uncertainties and disturbances) within a prescribed time. Then, a local control algorithm is designed by incorporating a sliding mode strategy such that the cluster TVFCT problem of the NRSs can be addressed within a prescribed time, where the convergence time can be set freely by choosing a tunable constant irrespective of the initial conditions. By constructing the Lyapunov function, several sufficient criteria for stability analysis are derived. Finally, some simulation examples are proposed to demonstrate the efficiency of the main results.
This paper proposes a fog weather data augmentation method for the unmanned surface vessels(USVs) via improved Generative Adversarial Network(GAN) model. First, a generator scheme for GAN is proposed with the guided g...
This paper proposes a fog weather data augmentation method for the unmanned surface vessels(USVs) via improved Generative Adversarial Network(GAN) model. First, a generator scheme for GAN is proposed with the guided generation of the atmospheric scattering model in this paper. A Laplacian Pyramid Based Depth Residuals model is added to the generator which reduces the difficulty of generating fog images caused by the degradation of water surface image and improves the quality of generated images. Finally, fog images are generated from sunny weather images collected with HUST-12C by LPBDR-GAN model and experiments show that generated images are very close to real fog images.
In recent decades, an avalanche of chaos-based cryptosystems have been proposed for information security. Most of these systems are not immune to the dynamical degradation of digital chaos and many have been shown to ...
详细信息
In recent decades, an avalanche of chaos-based cryptosystems have been proposed for information security. Most of these systems are not immune to the dynamical degradation of digital chaos and many have been shown to suffer from a lack of security. In this paper, a stream cipher system based on an analog-digital hybrid chaotic system is presented. The hybrid model can construct digital chaotic maps without degeneration and guarantee synchronization of analog chaotic systems for successful decryption. Moreover, focusing on the characteristics of low-dimensional chaotic maps, a modified three-dimensional Logistic map is proposed to improve the weaknesses of uneven distribution, low complexity and limited parameter space. Combining the three-dimensional Logistic map and the hybrid model, the proposed stream cipher has advantages of huge key space, virtually infinite cycle length and tight security. In particular, it is not affected by the dynamical degradation. Performance and security analyses indicate that the proposed stream cipher is highly resistant to various chaos-based attacks and cryptanalytic attacks.(c) 2021 Elsevier Inc. All rights reserved.
作者:
Xiang HuangHai-Tao ZhangSchool of Artificial Intelligence and Automation
the Engineering Research Center of Autonomous Intelligent Unmanned Systems the Key Laboratory of Image Processing and Intelligent Control and the State Key Laboratory of Digital Manufacturing Equipment and Technology Huazhong University of Science and Technology
The piezoelectric actuator is one kind of device that can drive nanoscale motion. However, the nonlinear hysteresis effect induced by its natural material greatly degrades its positioning accuracy. To handle this chal...
The piezoelectric actuator is one kind of device that can drive nanoscale motion. However, the nonlinear hysteresis effect induced by its natural material greatly degrades its positioning accuracy. To handle this challenging issue, this work develops a Koopman model predict control(Koopman-MPC) framework for the piezoelectric actuator. Specifically, the Koopman operator theory is adapted for modeling the piezoelectric actuator dynamics. A simple yet powerful linear model spanned in a high-dimensional space is thus constructed to characterize the hysteresis dynamics. Subsequently, upon the established Koopman model, an MPC scheme is put forward for tracking control of piezoelectric actuators. Therein, by sustained optimizing a cost function containing future outputs and control increments, the control input is obtained. Moreover, extensive tracking simulations are carried out on a simulated piezoelectric actuator for verifying the feasibility and effectiveness of the Koopman-MPC scheme.
This paper solves USV path planning problem constrained by multiple factors via ant-colony optimization ***, this paper uses the ways of multi-objective optimization to model the USV path planning problem. An improved...
This paper solves USV path planning problem constrained by multiple factors via ant-colony optimization ***, this paper uses the ways of multi-objective optimization to model the USV path planning problem. An improved ant colony algorithm named ACO-SA is put forward afterwards to effectively solve the problem. The algorithm is a combination of ACO algorithm(ant colony algorithm) and SA algorithm(simulated annealing algorithm), which has three improments: change the initial distribution of pheromone to guide the search when the algorithm has just started running;change the heuristic function and state transition probability taking three factors into consideration;change the pheromone update rule and make the ants compete for the right to update pheromone by simulated annealing algorithm, and update the best solution by the same ***, simulation experiment and field experiment are conducted to check the validity of ACO-SA algorithm.
The rapid evolution of Integrated Energy Systems (IESs) demands robust management of information transmission, which is critical for real-time monitoring, coordination, and operational efficiency. However, the increas...
详细信息
The rapid evolution of Integrated Energy Systems (IESs) demands robust management of information transmission, which is critical for real-time monitoring, coordination, and operational efficiency. However, the increasing complexity and costs associated with information exchange necessitate effective pricing mechanisms to ensure economic sustainability and optimal resource allocation. This paper presents an evolutionary game-theoretic framework to analyze regulatory strategies governing information transmission within IES. In the context of market dynamics, both market regulators and communication network operators are considered as actors with bounded rationality, emphasizing their strategic interplay within the system. The main contributions include formulating a model that treats communication network operators as independent entities, enhancing the understanding of IES member diversity and interactivity. This study introduces evolutionary game dynamics, providing new insights into optimizing regulatory policies. This paper also innovates by considering asset utilization in defining effective assets, potentially curbing excessive investment by communication network operators and preventing information transmission prices from soaring. A case study is provided to reveal the dynamics and equilibrium states of the regulatory game, offering theoretical support for refining regulatory strategies in IESs.
暂无评论