Ship detection based on multi-scale weighted fusion

Ship surveillance plays an important role in ensuring the safety of maritime transportation and navigation. Due to the influence of factors such as waves and special weather, the existing detection methods still cannot balance the accuracy, speed and the parameters of the model in the changeable and complex marine environment. To solve this problem, this paper proposes an improved real-time method based on YOLOv5, which has few parameters and achieves high detection accuracy with little memory and computation cost. Collaborative Attention (CA) mechanism is added to the network structure, which enables the model to more accurately locate and identify target regions. We also design a Spatial Pyramid Pooling module (SPP) and a weighted pyramid network called Bidirectional Feature Pyramid Network (BiFPN) based on the characteristics of the ships to better fuse feature information. Transformer encoder is introduced to capture long-distance dependencies and preserve global and local features to the greatest extent. Furthermore, the ability of our proposed structure to localize objects at each stage is improved through integrating the output of multiple modules. The experimental results show that, the comprehensive performance of this method is better than the existing technology in ship detection on different evaluation criteria.     

Ultraefficient Non-Doped Deep Blue Fluorescent OLED: Achieving a High EQE of 10.17% at 1000 cd m−2 with CIEy<0.08

The pursuit for efficient deep blue material is an ever-increasing issue in organic optoelectronics field. It is a long-standing challenge to achieve high external quantum efficiency (EQE) exceed 10% at brightness of 1000 cd m⁻² with a Commission International de L'Eclairage (CIE_y) <0.08 in non-doped organic light-emitting diodes (OLEDs). Herein, this study reports a deep blue luminogen, PPITPh, by bonding phenanthro[9,10-d]imidazole moiety with m-terphenyl group via benzene bridge. The non-doped OLED based on PPITPh exhibits an exceptionally high EQE of 11.83% with a CIE coordinate of (0.15, 0.07). The EQE still maintains 10.17% at the brightness of 1000 cd m⁻², and even at a brightness as high as 10000 cd m⁻², an EQE of 7.5% is still remained, representing the record-high result among non-doped deep-blue OLEDs at 1000 cd m⁻². The unprecedented device performance is attributed to the reversed intersystem crossing process through hot exciton mechanism. Besides, the maximum EQE of orange phosphorescent OLED with PPITPh as host is 32.02%, and remains 31.17% at the brightness of 1000 cd m⁻². Such minimal efficiency roll-off demonstrates that PPITPh is also an excellent phosphorescent host material. The result offers a new design strategy for the enrichment of high-efficiency deep blue luminogen.     

Event-triggered control for stochastic systems with multiple delays

This paper investigates the stabilization problem for continuous-time stochastic systems with multiple delays under continuous event-triggered mechanisms, of which both static case and dynamic case are considered individually. In order to avoid zeno phenomenon in every sample path, a suspension time after each successful execution is forced for our event-triggered mechanisms, resulting in intermittent detection of system states. Under such control strategy, we deduce mean square exponential stability of stochastic systems with multiple delays by means of Hanalay-type inequality and obtain a delay-dependent-based and less-conservative stabilization criterion without involving the upper bound of time delays. Besides, a co-design procedure is proposed for linear controller and event-triggered mechanisms. In the end, an illustrative example is presented to show effectiveness of the proposed co-design procedure and contrasts the system performance under static and dynamic event-triggered mechanisms.     

Adaptive neural network dynamic event-triggered control for strong interconnected stochastic nonlinear systems with output constraint

This paper proposes a dynamic event-triggered mechanism based command filtered adaptive neural network (NN) tracking control scheme for strong interconnected stochastic nonlinear systems with time-varying output constraints. By designing a state observer, the unmeasured states of the systems can be estimated. The NNs are utilized to handle the unknown intermediate functions. In the controller design process, the asymmetric time-varying barrier Lyapunov functions are used to guarantee that the systems outputs do not violate the constraint regions. By integrating the command filter with variable separation technique, the controller design process is more simple, and the problem of algebraic-loop can be solved which caused by interconnected functions. According to the Lyapunov stability theory, it can be ensured that all signals of the systems are bounded in probability. Finally, the availability of the developed control scheme can be showed by the simulation example.     

Web service adaptation: A decade’s overview

With the exponential growth of communication and information technologies, adaptation has gained a significant attention as it becomes a key feature of service-based systems, allowing them to operate and evolve in highly dynamic and uncertain environments. Although several Web service standards and frameworks have been proposed and extended, existing solutions do not provide a suitable architecture, in which all aspects of monitoring and adaptation (e.g., proactive, cross-layer, and autonomic adaptation) can be expressed. In addition, the emergence of new computing environments to host and execute various types of services (Web/cloud services, big data-intensive services, mobile services, microservices, etc.) raises the need for more efficient monitoring and adaptation systems. This survey aims to bring a synthesis and a road-map to the adaptation of service-based systems. We also discuss adaptation solutions in emerging service models, such as cloud services and big services. Based on an adaptation taxonomy which we extracted from the surveyed approaches, and by identifying the main requirements and goals of service adaptation in Web, cloud and big data environments, detailed analysis and discussions, as well as the open issues, are provided.     

融合双注意力机制的人群计数算法

针对背景复杂、遮挡、人群分布不均等人群计数常见问题,提出了一种结合联合损失的空间-通道双注意力机制卷积神经网络模型(joint loss-based space-channel dual attention network, JL-SCDANet).该网络前端进行图像粗粒度特征提取,中间加入空间注意力机制以及通道注意力机制突出图像重点区域,后端使用可加大感受野且不丢失图像分辨率的空洞卷积提取深层二维特征.此外,该模型结合联合损失函数进行训练,以增强模型的鲁棒性.为了验证模型的改进效果,在3个公共数据集(ShanghaiTech Part B、mall和UCF＿CC＿50)上分别进行了对比实验,在ShanghaiTech Part B数据集中平均绝对误差(MAE)和均方误差(MSE)分别达到了8.13和13.13;在mall数据集中MAE、MSE达到了1.78和2.28;在UCF＿CC＿50数据集中MAE、MSE分别达到了182.12和210.24,实验结果证明了该网络在提高人数统计准确率上的有效性.     

基于双分支融合注意力机制的图像分割算法

针对DeepLabV3+在特征提取阶段忽略了不同尺度特征重要程度出现的部分细节信息损失导致图像分割不细致,提出一种融合双分支特征提取和注意力机制的改进算法. ResNet101骨干网络初步提取出的特征图作为注意力机制的输入特征,解决了网络退化及梯度消失的问题,也能够捕获到被DeepLabV3+忽略的图像细节信息;设计双分支特征提取机制扩大特征提取能力,细化图像边缘信息以优化网络对不同尺度特征关注不均的问题;同时,联合采用交叉熵损失和类别不平衡函数两种损失函数作为损失函数,通过聚焦于前景样本降低背景的影响,提高算法分割精度.实验结果表明,改进算法在PASCAL VOC 2012和CityScapes数据集上的平均交并比(MIoU)值分别达到了79.92%和68.59%,与经典算法和基于DeepLabV3+改进的算法相比,特征提取的准确性有所提高,分割效果更优.     

基于MultiResUNet-SMIS的皮肤黑色素瘤图像分割

针对现有的皮肤黑色素瘤病灶分割精度不高的问题,结合现有卷积神经网络方法提出皮肤黑色素瘤图像分割方法 MultiResUNet-SMIS.首先,依据皮肤黑色素瘤成像特点,引入不同空洞率的空洞卷积替换普通卷积,在参数量相同的前提下扩大感受野,使网络模型能够适用于多尺度病灶分割任务;其次加入空间和通道注意力机制以重新分配特征权重,扩大感兴趣特征影响,抑制无关特征;最后融合Focal loss与Dice loss提出一种新的loss函数FD loss用于计算回归损失,解决前景背景像素不均衡问题,进一步提高网络模型的分割精度.实验结果表明,MultiResUNet-SMIS在ISIC-2018数据集上的Dice指数、IoU指数以及Acc准确率分别达到了89.47%、82.67%、96.13%,与原MultiResUNet以及UNet、UNet++、DeepLab V3+等主流方法相比, MultiResUNet-SMIS在皮肤黑色素瘤图像分割中具有更好的效果.     

基于YOLOv5的交通标志识别

交通标志识别是自动驾驶技术中的关键一部分.针对交通标志在道路场景中目标较小且识别精度较低的问题,提出一种改进的YOLOv5算法.首先在YOLOv5模型中引入全局注意力机制(GAM),提高网络捕获不同尺度交通标志特征的能力;其次将YOLOv5算法中使用的GIoU损失函数更换为更具回归特性的CIoU损失函数来优化模型,提高对交通标志的识别精度.最后在Tsinghua-Tencent 100K数据集上进行训练,实验结果表明,改进后的YOLOv5算法对交通标志识别的平均精度均值为93.00%,相比于原算法提升了5.72%,具有更好的识别性能.     

基于Attention-CGRU网络的中文语音情感识别

正确识别语音中包含的情感信息可以大幅提高人机交互的效率.目前,语音情感识别系统主要由语音特征抽取和语音特征分类两步组成.为了提高语音情感识别准确率,选用语谱图而非传统声学特征作为模型输入,采用基于attention机制的CGRU网络提取语谱图中包含的频域信息和时域信息.实验结果表明:在模型中引入注意力机制有利于减少冗余信息的干扰,并且相较于基于LSTM网络的模型,采用GRU网络的模型预测精确度更高,且在训练时收敛更快,与基于LSTM的基线模型相比,基于GRU网络的模型训练时长只有前者的60%.