摘要：全局滑模控制对系统参数变化和外部干扰的不敏感特性以及在响应的全过程都具有鲁棒性的特点,可消除直流无刷电机的非线性因素带来的不利影响,但是滑模控制在本质上的不连续开关特性会引起系统的抖振。为解决上述问题,首先,设计了一种改进型的全局滑模控制方法,通过优化滑模控制器的趋近律,提升了系统的趋近速率。然后,对边界层厚度动态变化的饱和函数进行了改进,有效减小了输出抖振,增强了系统的稳定性。最后,证明了该方法在李雅普诺夫函数意义下能够实现收敛稳定。通过仿真实验对比了所提改进型算法与全局滑模控制算法和传统滑模控制算法对无刷直流电机的控制效果,最终证明改进型全局滑模控制算法具有更小的抖振以及更好的快速性和鲁棒性。

摘要：近年来,基于深度学习的目标跟踪网络取得了显著的进展.这些网络主要采用两种类型的框架:双流双阶段框架和单流单阶段框架.然而,前者忽视了在特征提取过程中的信息交互,后者则受限于骨干网络自身的局限性.因此,本文采用独立骨干网络来直接构建跟踪器,并设计了一种轻量化的多尺度特征融合架构,以较低的计算成本增强了网络对多尺度信息的感知能力;同时,引入递归门控卷积作为特征学习单元,以自适应高阶空间交互实现了网络对特征的深层挖掘;此外,本文使用DropMAE预训练模型来进行网络初始化,以提升网络的泛化能力.实验结果表明,所提出的目标跟踪网络在多个大型跟踪数据集基准上都表现出优异的性能,并能以78.4 FPS的速度进行实时跟踪.

摘要：针对现有的基于深度学习的高光谱图像超分辨率重建方法无法通用于不同波段的高光谱图像以及缺乏可解释性等问题.提出一种融合光谱子空间映射和模型引导的高光谱图像超分辨率算法.首先,使用光谱子空间分解将原始图像映射到低维空间中,既可以增加光谱间的相关性又可以去除不同波段高光谱图像对网络的限制.其次,使用小波变换将稀疏矩阵分解为高频特征和低频特征,挖掘图像中的纹理和结构等高频信息.再者,以超分辨率重建模型为指导,将ADMM分解后的子模型优化展开为深度网络的形式,增加了深度网络设计的可解释性.最终,使用逆小波变换后将重建的系数矩阵映射到原始的全谱空间中.实验表明,提出的方法在定量指标和主观视觉方面均表现优异.

摘要：针对传统显微系统在覆盖较大面积的病理切片时需多次成像,并且运用图像拼接技术的挑战,设计了一款大视场变焦显微光学系统。通过巧妙结合大视场显微物镜和变焦筒镜,该系统实现了对病理切片的一次成像,从而获取了病理切片的全部信息。通过调整变焦筒镜系统的放大倍率,可以观察到局部疑似病变组织。使用ZEMAX光学软件分别对大视场显微物镜和变焦筒镜进行了优化设计。设计结果表明,组合系统焦距范围为90~800mm,实现12倍变焦,工作距离为100mm,分辨率可达到3.45μm,在全视场范围内,MTF在75lp/mm处均大于0.3,接近衍射极限。仿真结果表明镜头像质优良,最大物方视场达到18mm×18mm,成功实现了大面积的待测样品从细节到全貌的观测,显著提高了检测效率。

摘要：网络入侵检测是一种重要的网络安全方案.目前网络入侵检测模型都有较高精确度,但是模型复杂,参数量和计算量较大.针对该问题,设计了一种新的基于包粒度应用层负载的网络入侵检测一维卷积轻量模型.本文首先对UNSWNB15数据集的原始流量文件进行包粒度应用层负载数据提取,构造一维灰度特征向量.在此基础上,本文提出一种由新的一维深度可分离卷积残差模块组成,融入了全局上下文注意力机制(Global Context Attention Module)的一维卷积轻量模型Fast Payload,并进行了针对性的模型优化和可行性论证.Fast Payload模型在UNSWNB15数据集上的9分类任务中宏平均准确率达到82.433%,加权平均精确率达到90.820%,均高于对比模型;同时,该模型计算量和参数量均低于对比模型.其次本文提出了二阶段类别平衡损失函数GHM2StageLoss,有效解决了数据集的类别不平衡问题,相比其他类别平衡损失函数,效果更好.为方便后续研究的复现,本研究开源部分源代码,网址为https://***/sadantange/FastPayload.

摘要：近年来,相关研究已经证明能够通过强化学习生成稳健的双足机器人行走步态。然而,传统机器人的结构能效较低。针对这一问题,基于一种能效更高的伸缩腿结构的机器人,提出一种运用周期函数来交替奖励与惩罚双足触地的方法,来改进传统训练学习方法生成步态不规律的现象,鼓励智能体生成对称步态。实验证明,通过简单运用周期函数设计稠密奖励函数,可以有效指导伸缩腿机器人生成高效的运动模式,能够有效提高机器人的能效,且具有较高的鲁棒性。

摘要：传统的连续过程控制系统不能很好地实现工艺与控制相分离,生产效率低下。针对此问题,提出了一种基于改进调度器、选择器和同步器(scheduler selector synchronizer,S^(3))控制结构的批次过程控制系统架构。该架构在调度层中建立了基于G描述语言的组件描述和基于改进S^(3)控制结构的调度器,使工艺与控制相分离;在数据层中建立了相逻辑接口表;在控制层中基于ISA S88标准和IEC 61499标准建立了离散化的控制功能块,并设计了S^(3)控制结构的选择器和同步器实现工艺的可重构。实验结果表明,所提批次过程控制系统架构在不改变控制程序的情况下完成了不同工艺流程的控制任务,实现了工艺与控制相分离,并缩短了工艺流程修改时间。

摘要：在航空和汽车等领域,抬头显示技术(Head-Up Display, HUD)的应用日益广泛。鉴于现代社会用眼负荷增加,本研究针对青少年近视问题,提出了一种具有大视场角、低畸变特性的抬头显示系统设计方案。该系统设计考虑了8米远的投影距离,旨在通过减少用眼负担来预防近视的发展。同时考虑到自由曲面在光学设计中高度的设计自由度,以及其对畸变的有效控制,对系统进行了进一步优化。本文通过光学设计软件Code V进行设计与优化,所设计的两版系统在成像性能上均满足了既定要求,全视场角达到30˚,同时将畸变控制在5%以内。此外,本研究还涉及了样机的制造和实际体验测试,以验证所提出设计方案的可行性。通过实际应用测试,进一步证实了该抬头显示系统设计的有效性和实用性,为未来相关技术的发展和应用提供了科学依据和实践指导。In the fields of aviation and automobiles, the application of Head-Up Display (HUD) is becoming more and more extensive. In view of the increasing eye load in modern society, this study proposes a design scheme of a head-up display system with a large field of view and low distortion characteristics for the problem of juvenile myopia. The system design takes into account the projection distance of 8 meters away, aiming to prevent the development of myopia by reducing the burden on the eye. Considering the high degree of design freedom of freeform surface in optical design and its effective control of distortion, the system is further optimized. In this paper, the optical design software Code V is used for design and optimization. The two designed versions of the system meet the established requirements in terms of imaging performance. The full field of view angle is up to 30˚, and the distortion is controlled within 5%. In addition, this study also involves prototype manufacturing and actual experience testing to verify the feasibility of the proposed design. Through the practical application test, the effectiveness and practicability of the design of the head-up display system are further confirmed, which provides a scientific basis and practical guidance for the development and application of related technologies in the future.

摘要：精准测量满载和卸料时颗粒物质对筒仓内壁的法向应力,掌握应力随时间和高度的变化规律,与筒仓的安全运行密切相关。本文设计了一种基于全桥电路的电阻应变式传感器,通过搭建筒仓应力测量系统,采用均值粒径为1.29 mm的透明玻璃珠进行实验,测量筒仓静态堆积和卸料过程中的壁面法向应力。静态应力的测量结果表明,颗粒填充完成后,在圆柱形部分出现了静态应力随着高度的降低而减小的区域,这与Janssen模型不一致。然后我们分析了每一个检测点在卸料过程中动态应力最大值的时空分布,结果表明,动态应力最大值出现在圆柱形中下部分,而不是筒仓几何转变处。最后我们计算了最大动态应力与静态应力的比值,得到了每个检测点的超压系数,发现超压系数最大值出现在H = 215 mm处。该研究对筒仓结构设计具有重要意义,有助于对应力增幅显著的区域进行结构优化,为提高筒仓的安全性和耐久性提供了关键参考。Accurately measuring the normal pressure exerted by granular materials on the inner walls of silos during loading and unloading, and understanding the variation of pressure over time and height, is closely related to the safe operation of silos. This study designed a resistive strain gauge sensor based on a full-bridge circuit and constructed a silo pressure measurement system. Experiments were conducted using transparent glass beads with an average particle size of 1.29 mm to measure the normal pressure on the silo walls during static filling and unloading processes. The results of static pressure measurements showed that, after filling, a region in the cylindrical section exhibited a decrease in static pressure with decreasing height, which is inconsistent with the predictions of the Janssen model. Subsequently, we analyzed the spatiotemporal distribution of the maximum dynamic pressure at each measurement point during unloading. The results showed that the maximum dynamic pressure occurred in the lower-middle part of the cylindrical section rather than at the geometric transition of the silo. Finally, we calculated the ratio of the maximum dynamic pressure to the static pressure at each measurement point to obtain the overpressure coefficient, and found that the maximum overpressure coefficient appeared at H = 215 mm. This study is of great significance to the structural design of silos, as it helps optimize the structure in regions with significant pressure amplification during unloading, providing key references for improving the safety and durability of silo operations.

摘要：从单张野外图片生成高保真3D人脸纹理是一项具有挑战的工作,现有方法在颜色和光照恢复方面已经取得了显著的进展,但是却仍然无法较好地重建中高频纹理细节。其主要原因在于真实面部UV纹理数据集的匮乏,现有模型大多基于合成UV纹理图训练模型,由于缺少真实标签的监督其与真实UV纹理相比必然存在较大的差异,从而导致模型学习到错误的纹理分布。基于以上思考,我们尝试使用原始图片空间中的细节纹理来引导UV空间中UV纹理图的生成,并提出两阶段训练方式以缓解仅使用合成UV纹理图训练模型带来的个性化细节缺失问题。此外,借助于扩散模型在图像生成任务中的卓越性能,我们还设计了一款跨域引导扩散模型,其将空间域和频率域中的细节信息编码为高级语义条件,用来引导扩散模型的生成过程,从而实现近乎精确的重建。最后,我们将跨域引导扩散模型作为UV纹理生成器嵌入到三维重建框架中,用于重建高保真的3D人脸纹理。实验结果表明本文提出的跨域引导扩散模型能较好地生成中高频纹理细节,并在定量和定性分析中明显优于其他3D人脸纹理生成工作。Generating high-fidelity 3D facial textures from a single outdoor image is a challenging task. While existing methods have made significant progress in color and lighting recovery, they still struggle to accurately reconstruct mid-to-high frequency texture details. This is primarily due to the lack of real facial UV texture datasets. Most models are trained using synthetic UV texture maps, which inherently differ from real UV textures due to the absence of ground truth supervision, leading to inaccurate texture distribution learning. In light of this, we attempt to use detailed textures from the original image space to guide the generation of UV texture maps in the UV space. We propose a two-stage training approach to alleviate the loss of personalized details caused by training models solely on synthetic UV texture maps. Additionally, leveraging the exceptional performance of diffusion models in image generation tasks, we design a cross-domain guided diffusion model. This model encodes detailed information from both spatial and frequency domains into high-level semantic conditions to guide the diffusion process for near-accurate reconstruction. Finally, we integrate the cross-domain guided diffusion model as a UV texture generator into a 3D reconstruction framework to reconstruct high-fidelity 3D facial textures. Experimental results demonstrate that our proposed cross-domain guided diffusion model effectively generates mid-to-high frequency texture details and significantly outperforms other 3D facial texture generation metho