基于改进YOLOv7的小目标检测

目前的目标检测技术已趋于成熟,但小目标检测仍是研究的难点。针对目标检测过程中小目标检测更容易出现漏检等问题,提出一种改进的YOLOv7目标检测模型。结合特征分离合并思想,对YOLOv7网络模型中的MPConv模块进行改进,以减少网络特征处理过程造成的特征损失,并通过实验确定放置改进MPConv模块的最佳位置。由于小目标检测过程中容易出现漏检的现象,利用ACmix注意力模块提高网络对小尺度目标的敏感度,降低噪声所带来的影响。在此基础上,使用SIoU替换原YOLOv7网络模型中的CIoU来优化损失函数,减少损失函数自由度,提高网络鲁棒性。在Okahublot公开的FloW-Img子数据集上进行实验,结果表明,对于数据集中的密集、小目标和超小目标三种情况的图片,改进后的YOLOv7网络模型相比原网络,漏检情况得到明显改善,且mAP达到71.1%,相比基线YOLOv7网络模型提升了4个百分点,检测效果优于原网络模型与传统经典目标检测网络模型。     

An Artificial Autonomic Nervous System That Implements Heart and Pupil as Controlled by Artificial Sympathetic and Parasympathetic Nerves

The autonomic nervous system maintains homeostasis in organisms through complex neural pathways and responds adaptively to changes in the external and internal environment. The fabrication of an artificial autonomic nervous system is reported that replicates combined effects of sympathetic and parasympathetic nerves on cardiac activity and pupillary control, to mimic the regulation of autonomic nervous system to external changes. The artificial autonomic nerve-controlled pupil contraction and relaxation, modulating the rate of heartbeats for normal cardiac rhythm and arrhythmia as reflected by blink rates of a signal indicator. These functions are switched by using a parallel-channeled synaptic transistor with a special n-i-p heterostructure that has a 2D h-BN insulator in the middle to provide barrier against ion injection into the 2D MoS₂ bottom n-channel and enable short-term plasticity as induced by acetylcholine, and the electrochemical doping reaction occurred at the P3HT nanowire p-channels on top to enable relatively long-term plasticity as induced by noradrenaline. Low-energy consumption down to femtojoule and an ultrahigh paired-pulse facilitation index up to ≈455% are demonstrated. An artificial neural network based on device characteristics achieves a high recognition accuracy for electrocardiogram patterns. This study extends insights into artificial nerves-inspired biological signal processing and recognition.     

面向软件供应链的异常分析方法综述

软件在国民经济的各个领域占据越来越重要的地位.万物互联的大背景下,信息之间的交互、分析、协同变得越来越普遍,程序/软件之间的依赖关系逐渐增多,这使得人们对系统可靠性和健壮性提出了更高的要求.由开源组件和第三方组件构成的软件供应链,其所面临的安全问题近年来成为了学术界和工业界共同关注的焦点.库函数作为开源软件的重要组成部分,与软件供应链安全有着密切的联系.为了提高软件开发效率,软件库或应用程序编程接口(API)在程序编写过程中会被频繁使用,但库函数中存在的错误或漏洞可能会被攻击者利用,从而损害软件供应链安全.这些错误或漏洞往往与库函数中存在的异常有关,因此本文对适用于库函数的异常分析方法从精度和效率两方面分别进行总结归纳,对于每种异常分析方法的基本思想和重要过程进行阐述,并针对库函数异常分析面临的挑战给出了初步解决思路.对软件供应链中的库函数进行异常分析有助于增强软件系统的健壮性,进而保障软件供应链的安全.     

基于模型预测—中枢模式发生器的六足机器人轨迹跟踪控制

为了模仿动物卓越的运动能力和环境适应能力,提出了六足仿生机器人的轨迹跟踪控制方法。首先建立了机器人的运动学模型,接着通过转向参数将机器人的速度和角速度与中枢模式发生器（CPG）参数结合起来,设计了转换函数。然后通过转换函数将模型预测控制器和CPG网络结合起来,提出了基于CPG的模型预测控制器（MPC-CPG）,并证明了其稳定性。最后对机器人跟踪圆周轨迹和直线轨迹进行了仿真和实验。实验表明,在有初始误差的条件下,机器人在MPC-CPG控制器的作用下能够快速地消除位置误差和航向角误差,跟踪上参考轨迹。轨迹跟踪的位置误差始终保持在-0.1～0.1 m,航向角误差保持在-27?～20?。在MPC-CPG控制器的作用下,机器人不仅具有较高的轨迹跟踪精度,同时还表现出良好的运动平滑性和协调性,进一步验证了所提出的MPC-CPG控制器的有效性。     

结合Transformer与非对称学习策略的图像检索

目的 图像检索是计算机视觉领域的一项基础任务,大多采用卷积神经网络和对称式学习策略,导致所需训练数据量大、模型训练时间长、监督信息利用不充分。针对上述问题,本文提出一种Transformer与非对称学习策略相结合的图像检索方法。方法 对于查询图像,使用Transformer生成图像的哈希表示,利用哈希损失学习哈希函数,使图像的哈希表示更加真实。对于待检索图像,采用非对称式学习策略,直接得到图像的哈希表示,并将哈希损失与分类损失相结合,充分利用监督信息,提高训练速度。在哈希空间通过计算汉明距离实现相似图像的快速检索。结果 在CIFAR-10和NUS-WIDE两个数据集上,将本文方法与主流的5种对称式方法和性能最优的两种非对称式方法进行比较,本文方法的mAP(mean average precision)比当前最优方法分别提升了5.06%和4.17%。结论 本文方法利用Transformer提取图像特征,并将哈希损失与分类损失相结合,在不增加训练数据量的前提下,减少了模型训练时间。所提方法性能优于当前同类方法,能够有效完成图像检索任务。     

Robust integrated sequential covariance intersection fusion Kalman filters and their convergence and stability for networked sensor systems with five uncertainties

For networked sensor systems (NSSs) with hard and soft sensors including five uncertainties, two universal approaches of solving the robust fusion estimation problems are presented. It includes an integrated sequential covariance intersection (SCI) fusion minimax robust Kalman filtering approach with cross-covariance information and a generalized Lyapunov equation approach with four pairs of Lyapunov equations. Applying them, the robust local and SCI fused time-varying and steady-state Kalman filters are presented in the sense that their actual estimation error variances are guaranteed to have the corresponding minimal upper bounds. The equivalent batch SCI fusers are also presented. Their robustness and accuracy relations are proved, and the sensitivity of the SCI fuser with respect to the fused orders of sensors is analyzed. Applying the dynamic error system analysis method and the dynamic variance error system analysis method, a new convergence and absolute asymptotic stability theory of robust fusion Kalman filtering is presented. The classical Kalman filtering convergence and stability theory is developed. Compared with the original covariance intersection fuser, they significantly reduced the computational complexity and burden. Compared with the optimal and conservative SCI fusers, they significantly improved the robust accuracies. They are suitable to deal with asynchronous or random delayed data and are suitable for real-time applications. A simulation applied to the two-mass spring damper mechanical system shows their effectiveness.     

Optimal reduced-order interval observer design for uncertain continuous-time linear systems

This article investigates the reduced-order interval observer (R-IO) design technique for continuous-time linear systems with unknown external disturbances and measurement noises. First, we propose a coupled R-IO structure with more design degrees of freedom, and it can be directly applied not only to solve the difficulty of the error system cooperativity construction but to relax the constraint on the sensor measurement noises. Second, the R-IO existence condition is derived as a set of matrix equations (MEs), and a complete solution, explicitly showing the available design parameters, to such an R-IO is further obtained by solving the MEs. Third, using the solution, an integrated optimization indicator of the R-IO performance is built as the valid selection mechanism of these parameters. Finally, the efficiency of the obtained results is illustrated through a numerical example and a practical example.     

基于相机感知距离矩阵的无监督行人重识别

在跨场景、跨设备的行人重识别中虽然增加了可利用的行人数据,但由于行人姿态不同、部分遮挡现象,难以避免引入样本噪声,在聚类过程中易生成错误的伪标签,造成标签噪声,影响模型的优化。为减弱噪声影响,应用相机感知的距离矩阵对抗相机偏移引起的样本噪声问题,利用对噪声鲁棒的动态对称对比损失减少标签噪声,提出基于相机感知距离矩阵的无监督行人重识别算法。在聚类前通过更改度量行人特征相似度的距离矩阵,利用相机感知距离矩阵来增强类内距离度量准确性,减少由于拍摄视角不同对聚类效果造成的负面影响。同时,结合噪声标签学习方法,进行损失设计,提出动态对称对比损失函数,联合损失训练,不断精炼伪标签。在DukeMTMC-reID和Market-1501两个数据集上进行实验,验证了提出方法的有效性。     

Optimized tracking control using reinforcement learning strategy for a class of nonlinear systems

This paper is to develop a simplified optimized tracking control using reinforcement learning (RL) strategy for a class of nonlinear systems. Since the nonlinear control gain function is considered in the system modeling, it is challenging to extend the existing RL-based optimal methods to the tracking control. The main reasons are that these methods' algorithm are very complex; meanwhile, they also require to meet some strict conditions. Different with these exiting RL-based optimal methods that derive the actor and critic training laws from the square of Bellman residual error, which is a complex function consisting of multiple nonlinear terms, the proposed optimized scheme derives the two RL training laws from negative gradient of a simple positive function, so that the algorithm can be significantly simplified. Moreover, the actor and critic in RL are constructed by employing neural network (NN) to approximate the solution of Hamilton–Jacobi–Bellman (HJB) equation. Finally, the feasibility of the proposed method is demonstrated in accordance with both Lyapunov stability theory and simulation example.     

一种改进残差深度网络的多目标分类技术

由于受场景、视角、光照、尺度变化以及局部变形等因素的影响,对重叠目标、拥挤目标、小目标的识别精度较低,提出了一种改进多支路的残差深度卷积神经网络来提高多目标识别的准确度。首先,在第一个卷积残差块layer1后保留恒等映射的同时,增加一个1×1的短接分支尽可能多的保留原始特征;再平行嵌入一个修改激活函数RELU6的空间_通道注意力机制模块（CBAM）;最后这三个特征图进行融合。融合后的特征层着重关注空间和通道中比较显著的信息,从而增强特征图的特征表达能力,以至于卷积神经网络(CNN)获得更多的判别特征,从而大大提高物体识别精度。在FashionMNIST和Cifar10两个数据集的对比性实验显示改进的resnet50算法是准确性-速度较为折中的目标识别模型。