摘要：
Object co-segmentation is a challenging task, which aims to segment common objects in multiple images at the same time. Generally, common information of the same object needs to be found to solve this problem. For various scenarios, common objects in different images only have the same semantic information. In this paper, we propose a deep object co-segmentation method based on channel and spatial attention, which combines the attention mechanism with a deep neural network to enhance the common semantic information. Siamese encoder and decoder structure are used for this task. Firstly, the encoder network is employed to extract low-level and high-level features of image pairs. Secondly, we introduce an improved attention mechanism in the channel and spatial domain to enhance the multi-level semantic features of common objects. Then, the decoder module accepts the enhanced feature maps and generates the masks of both images. Finally, we evaluate our approach on the commonly used datasets for the co-segmentation task. And the experimental results show that our approach achieves competitive performance. (C) 2020 Elsevier B.V. All rights reserved.
期刊：
International Journal of Future Generation Communication and Networking,2013年6(3):1-12 ISSN：2233-7857
作者机构：
[Duan, Jia-Qi; Li, Shining] Shaanxi Key Lab of Embedded System Technology, School of Computer Science, Northwestern Polytechnical University, Xi'an, 710129, China;[Ning, Guoqin] Department of Information Technology, Huazhong Normal University, Wuhan, 430079, China
摘要：
Cognitive radio enabled vehicular networks (CR-VNETs) is a new communication paradigm that enables moving vehicles to identify spectrum opportunities along busy streets and freeways. This detected spectrum may possibly lie in licensed frequency bands, and can be used for emergency communications, such as by primary responders during crises events. Spectrum sensing ensures that this spectrum is not currently occupied by licensed users, who have priority access rights. However, as the vehicles are in motion, the spectrum sensing at a given location must be completed with minimum delay, a challenge for classical energy and feature based detection schemes. This paper presents a new distributed compressive sampling technique that allows individual vehicles to report partial information to a centralized base station (BS), with an overhead of only few bytes. Thus, we tradeoff reporting time with processing complexity at the BS, which is tasked with re-constructing the overall spectrum utilization from these portions. Simulation results reveal significant improvements in detection time and accuracy, making our approach suitable for CR-VNETs.
作者机构：
[Ning, Guoqin] Cent China Normal Univ, Dept Informat Technol, Wuhan, Peoples R China.
通讯机构：
[Ning, Guoqin] C;Cent China Normal Univ, Dept Informat Technol, Wuhan, Peoples R China.
会议地点：
Paris, FRANCE
会议主办单位：
[Ning, Guoqin] Cent China Normal Univ, Dept Informat Technol, Wuhan, Peoples R China.
会议论文集名称：
IEEE Wireless Communications and Networking Conference
关键词：
centralized cognitive radio networks;mobile sensing;maximum likelihood estimator
摘要：
Cognitive radio (CR) networks rely on the spectrum sensing function to ensure that there is no interference to the licensed or primary users (PUs). Typically, sensing algorithms assume a static PU activity model, i.e., spectrum usage model, which is constant for a given channel and known in advance. This approach fails to capture the dynamic and time-varying behavior of the PUs. In this paper, a spectrum usage detection approach based on time prediction for centralized CR networks is proposed. The proposed approach allows the CR users to learn about the activity of the PUs, and adapt to subsequent changes. CR base station selects CR user with the longest sensing time predicted by a mobile model. Each selected mobile CR user uses maximum likelihood estimator (MLE) on the observed ON/OFF period samples to estimate the average busy and idle periods. In addition, CR base station employs mean square error (MSE) to determine when the fine sensing should stop, and exploits the variation of MSE to restart the fine sensing. Simulation results reveal that our proposed method can efficiently and quickly track the dynamics of the PU spectrum usage.
作者机构：
[Ning, Guoqin; Qiu, Duo; Su, Jian] Department of Information Technology, Huazhong Normal University, Wuhan, China;[Ning, Guoqin; Duan, Jiaqi] Department of Electrical and Computer Engineering, Northeastern University, Boston, United States
通讯机构：
[Ning, Guoqin] H;Huazhong Normal Univ, Dept Informat Technol, Wuhan, Peoples R China.
会议名称：
第三届IEEE无线通讯、网络技术暨移动计算国际会议
会议时间：
2007-09-21
会议地点：
上海
会议论文集名称：
第三届IEEE无线通讯、网络技术暨移动计算国际会议论文集
关键词：
heterogeneous hierarchical wireless networks;handoff;quality of service;resource preemption
摘要：
A handoff scheme based on guard channel, queuing and channel preemption (GCQCP) is proposed to guarantee the QoS of handoff traffic in the heterogeneous hierarchical wireless networks. On the basis of bidirectional call-overflow between different networks, guard channels are set for real-time traffic handoffs and buffer queues are set for nonreal-time traffic handoffs. In order to further decrease the call dropping probability of real-time traffic handoffs, channel preemption is also used, which principle is that real time traffic handoff calls can preempt the channel resources occupied by the nonreal-time traffic. Simulation results show that the proposed handoff control scheme can remarkably decrease the dropping probability of real-time traffic handoffs, the channel utilization of overall heterogeneous system is slightly increased simultaneously.
作者机构：
[宁国勤; 彭烈新; 朱光喜; 卢小峰] Wuhan National Laboratory of Optoelectronics, Huazhong University of Science and Technology, Wuhan 430074, China;Department of Information Technology, Huazhong Normal University, Wuhan 430079, China
通讯机构：
Wuhan National Laboratory of Optoelectronics, Huazhong University of Science and Technology, China
