排序方式: 共有131条查询结果,搜索用时 31 毫秒
1.
Xin Xu Xiaolong ZhangAuthor VitaeHaidong FuAuthor Vitae Li ChenAuthor VitaeHong ZhangAuthor Vitae Xiaowei FuAuthor Vitae 《Computers & Electrical Engineering》2014
A robust autofocus system is a ubiquitous function in today’s mobile phone camera applications. However, due to the power consumption and size requirements, it is difficult for the autofocus function to be implemented into the design of mobile phone cameras. This paper presents a passive autofocus system with low computational complexity. This system uses a novel contrast measurement to determine degree of image sharpness, which can better reflect the information about image discontinuities. In order to gauge the performance of this measurement, a modified peak search strategy is used in the experiments. The experimental results from several typical image sequences validate the effectiveness of the proposed method. 相似文献
2.
3.
异构车载网络环境下如何选择接入网络对于车载终端用户的服务体验而言至关重要,目前基于Q学习的网络选择方法利用智能体与环境的交互来迭代学习网络选择策略,从而实现较优的网络资源分配.然而该类方法通常存在状态空间过大引起迭代效率低下和收敛速度较慢的问题,同时由于Q值表更新产生的过高估计现象容易导致网络资源利用不均衡.针对上述问题,基于多智能体Q学习提出一种适用于融合5G通信异构车载网络的选择方法M QSM.该方法采用多智能体协作学习的思想,利用双Q值表交替更新的方式来获得动作选择的总回报值,最终实现异构车载网络环境下长期有效的最优网络切换决策集合.实验结果表明,与同类型方法相比较,M QSM在系统总切换次数、平均总折扣值和网络容量利用率方面表现出更好的性能. 相似文献
4.
针对已有的航运监控图像识别模型C3D里中级表征学习能力有限,有效特征的提取容易受到噪声的干扰,且特征的提取忽视了整体特征与局部特征之间关系的问题,提出了一种新的基于注意力机制网络的航运监控图像识别模型。该模型基于卷积神经网络(CNN)框架,首先,通过特征提取器提取图像的浅层次特征;然后,基于CNN对不同区域激活特征的不同响应强度,生成注意力信息并实现对局部判别性特征的提取;最后,使用多分支的CNN结构融合局部判别性特征和图像全局纹理特征,从而利用局部判别性特征和图像全局纹理特征的交互关系提升CNN学习中级表征的能力。实验结果表明,所提出的模型在航运图像数据集上的识别准确率达到91.8%,相较于目前的C3D模型提高了7.2个百分点,相较于判别滤波器组卷积神经网络(DFL-CNN)模型提高了0.6个百分点。可见所提模型能够准确判断船舶的状态,可以有效应用于航运监控项目。 相似文献
5.
重轨的平直度检测是重轨生产过程中重要的一个环节,该项检测的速度、准确性、稳定性等因素直接影响重轨总体质量。 本文以中国某大型钢铁集团轨梁厂的重轨平直度检测现状为背景,利用激光传感、数据采集、计算技术等现代技术,设计了一套基于并行计算的重轨平直度在线检测系统。系统采用14个激光位移传感器和2个旋转编码器分别对于重轨的7个水平方向位移测点、7个竖直方向位移测点、和重轨传输方向的位移进行实时测量,得到的信号转换为数字信号后,在PC机内通过GPU并行计算进行同步配准、块内数据分析、块间数据分析、帧间数据分析等处理,除去粗糙度、氧化铁皮、高频振动、低频振动等干扰,最后得到并输出平直度信息。实验及现场应用表明,该方法检测速度可达到3m/s、精确度可达到0.1mm. 相似文献
6.
在直接处理点云的三维神经网络中,采样阶段实现了对原始点云中关键点的筛选,对于整个网络的性能及网络的抗噪能力具有重要作用。目前主流的最远点采样(FPS)方法在处理大规模3D点云数据时计算量大且耗时,并且低采样率时经过FPS采样后模型性能下降明显。针对这两个问题,提出一种面向低采样率的点云数据处理网络AS-Net。设计一个新的采样模块代替原backbone中的FPS,其由两个Layer组成,每个Layer基于长短期记忆网络获取原始点云与采样点云之间的联系权重,从而高效提取关键信息,去除冗余信息。在此基础上,利用注意力机制选择特征值较高的原始点云作为采样点,采样点作为后序任务的关键点输入到网络,进一步提高网络模型性能。基于ModelNet40数据集的实验结果表明,在低采样率条件下,AS-Net仍可达到81.6%的分类准确率,与使用FPS作为采样方法的网络模型相比提高52.7%。此外,其对噪声干扰具有很强的鲁棒性,对于大场景的分割时间效率优于同类采样方法。 相似文献
7.
One of the major expenses for steel structures is the anti-corrosion maintenance tasks. The maintenance of a steel structure depends on regular inspections, and visual inspections are often adopted in Taiwan. Using the naked eye to determine the rusted area percentage greatly depends on the experience of the inspector, resulting in subjective results. As an alternative, an algorithm consisting of three different approaches is proposed to automatically process images. The Hue percentage and coefficient of variation (COV) of the gray levels are used to divide images into three groups in which each group is assessed using a specific recognition technique. The three proposed techniques are the following: the traditional K-means method in the H component, the double-center-double-radius (DCDR) algorithm in the Red-Green-Blue (RGB) color space and DCDR in the Hue-Saturation-Intensity (HSI) color space. Additionally, the Least Square Support Vector Machine (LS-SVM) was adopted to predict the radii in the DCDR approaches. One hundred images, mostly collected outdoors, were used to verify the proposed algorithm. Promising performance was observed, particularly for images with non-uniform illumination. 相似文献
8.
More and more cores are integrated onto a single chip to improve the performance and reduce the power consumption of CPU without the increased frequency. The cores are connected by lines and organized as a network, which is called network on chip (NOC) as the promising paradigm of the processor design. However, it is still a challenge to enhance performance with lower power consumption. The core issue is how to map the tasks to the different cores to take full advantages of the on-chip network. In this paper, we proposed a novel mapping algorithm with power-aware optimization for NOC. The traffic of the tasks will be analyzed. The tasks of the same application with high communication with the others will be mapped to the on-chip network as neighborhoods. And then the tasks of different applications are mapped to the cores step by step. The mapping of the tasks and the cores is computed at run-time dynamically and implement online. The experimental results showed that this proposed algorithm can reduce the power consumption in communication and the performance enhanced. 相似文献
9.
《Parallel Computing》2013,39(10):567-585
We examine the problem of reliable workflow scheduling with less resource redundancy. As scheduling workflow applications in heterogeneous systems, either for optimizing the reliability or for minimizing the makespan, are NP-Complete problems, we alternatively find schedules for meeting specific reliability and deadline requirements. First, we analyze the reliability of a given schedule using two important definitions: Accumulated Processor Reliability (APR) and Accumulated Communication Reliability (ACR). Second, inspired by the reliability analysis, we present three scheduling algorithms: RR algorithm schedules least Resources to meet the Reliability requirement; DRR algorithm extends RR by further considering the Deadline requirement; and dynamic algorithm schedules tasks dynamically: It avoids the “Chain effect” caused by uncertainties on the task execution time estimates, and relieves the impact from the inaccuracy on failure estimation. Finally, the empirical evaluation shows that our algorithms can save a significant amount of computation and communication resources when performing a similar reliability compared to Fault-Tolerant-Scheduling-Algorithm (FTSA) algorithm. 相似文献
10.