针对水下自主航行器(AUV)在复杂水下环境中进行三维路径规划时,沙猫群算法所面临的障碍物规避能力有限、迭代效率较低以及容易陷入局部最优解等问题,本研究提出了一种将沙猫群优化算法与莱维飞行方法相结合的策略,旨在提升沙猫群算法的...针对水下自主航行器(AUV)在复杂水下环境中进行三维路径规划时,沙猫群算法所面临的障碍物规避能力有限、迭代效率较低以及容易陷入局部最优解等问题,本研究提出了一种将沙猫群优化算法与莱维飞行方法相结合的策略,旨在提升沙猫群算法的整体性能。基于混沌映射的均匀分布特性,改进初始种群的生成策略,有效增强了群体的多样性;此外,引入互利共生机制,并结合莱维飞行策略进行调整,显著提高算法寻找全局最优解的能力。这一改进不仅提高了算法的收敛速度,也提升了求解精度。通过静态障碍与动态洋流干扰场景的仿真测试,改进的沙猫群算法(LVSCSO)在全局收敛性上显著优于PSO、GA等六类算法:最优解偏离度降低21.4%,最差解稳定性提升33.7%,平均解精度优化19.5%。结果表明,LVSCSO可有效应对复杂水下路径规划任务(如海底勘探),具备工程部署潜力。For underwater autonomous vehicle (AUV) in complex underwater environment for 3D path planning, the sand cat group algorithm facing obstacle avoidance ability, slow convergence and easily into local optimal solution, this study puts forward a sand cat group optimization algorithm and levy flight method combining strategy, aims to improve the overall performance of the sand cat group algorithm. By initializing the initial population with the consistency of chaotic mapping, the population diversity is effectively enhanced. In addition, the mutualism mechanism and the adjustment of Levy flight strategy significantly enhance the algorithm’s ability to find the global optimal solution. This improvement not only improves the convergence speed of the algorithm, but also improves the solution accuracy. Through simulation tests in scenarios of static obstacles and dynamic current interference, the improved Sand Cat Swarm Optimization algorithm (LVSCSO) significantly outperforms six types of algorithms including PSO and GA in terms of global convergence: the deviation of the optimal solution is reduced by 21.4%, the stability of the worst solution is improved by 33.7%, and the average solution accuracy is optimized by 19.5%. The results indicate that LVSCSO can effectively address complex underwater path planning tasks (such as seabed exploration) and has potential for engineering deployment.展开更多
低剂量CT技术在显著降低患者辐射剂量的同时,不可避免地引入多样化的噪声与伪影,其强度与分布特性因成像条件而异,对图像质量及临床诊断准确性构成严峻挑战。传统图像去噪方法通常基于先验知识构建数学模型,虽能有效抑制部分噪声,但其...低剂量CT技术在显著降低患者辐射剂量的同时,不可避免地引入多样化的噪声与伪影,其强度与分布特性因成像条件而异,对图像质量及临床诊断准确性构成严峻挑战。传统图像去噪方法通常基于先验知识构建数学模型,虽能有效抑制部分噪声,但其优化过程依赖人工参数调谐,存在计算复杂度高、图像细节保留不足等固有缺陷。近年来,基于深度学习的去噪方法凭借其强大的非线性特征提取与端到端优化能力,在处理复杂噪声场景时展现出显著优势。本文系统性地介绍了低剂量CT图像去噪领域的研究进展:首先剖析了传统方法的理论框架及其局限性;随后重点探讨深度学习方法的技术原理、代表性模型架构及其在医学影像中的创新应用;最后,总结当前技术面临的核心挑战,并展望未来研究方向,旨在为低剂量CT成像技术的优化与临床转化提供理论依据与技术参考。While significantly reducing patient radiation exposure, low-dose CT technology inevitably introduces diverse noise and artifacts, whose intensity and distribution characteristics vary with imaging conditions, posing a serious challenge to image quality and clinical diagnostic accuracy. Traditional image denoising methods, typically based on prior knowledge to construct mathematical models, can effectively suppress some noise. However, their optimization process relies on manual parameter tuning, exhibiting inherent limitations such as high computational complexity and insufficient preservation of image details. In recent years, deep learning-based denoising methods have demonstrated significant advantages in handling complex noise scenarios, leveraging their powerful nonlinear feature extraction and end-to-end optimization capabilities. This paper systematically introduces the research progress in the field of low-dose CT image denoising: first, it analyzes the theoretical frameworks and limitations of traditional methods;then, it focuses on the technical principles of deep learning methods, representative model architectures, and their innovative applications in medical imaging;finally, it summarizes the core challenges currently faced by the technology and outlines future research directions, aiming to provide theoretical foundations and technical references for the optimization and clinical translation of low-dose CT imaging technology.展开更多
基于深度学习的推荐算法逐渐成为推荐系统领域的主流研究方向。然而,大多数现有工作仅基于单一的用户与物品交互数据,并且缺乏可解释性。本文对用户评论进行充分挖掘,并且额外引入物品信息来缓解冷启动问题并提高推荐算法的准确性。该...基于深度学习的推荐算法逐渐成为推荐系统领域的主流研究方向。然而,大多数现有工作仅基于单一的用户与物品交互数据,并且缺乏可解释性。本文对用户评论进行充分挖掘,并且额外引入物品信息来缓解冷启动问题并提高推荐算法的准确性。该算法利用BERT预训练模型来处理文本数据,并将用户与物品的评论特征与矩阵分解得到的潜在特征相融合,最后在评分预测任务中使用Kolmogorov-Arnold网络进行特征学习。通过本文算法与其他基线算法在亚马逊评论数据集上进行对比,该算法与基线算法相比显著提高了评分预测的精度以及准确率和召回率。本研究通过深入挖掘用户评论文本和物品描述信息,证明其在提升推荐算法准确性方面的显著效果,为推荐系统的研究提供了新的思路。Recommendation algorithms based on deep learning have emerged as a prominent research in the field of recommender systems. However, most existing approaches rely solely on user-item interaction data and lack interpretability. This article thoroughly explores user reviews and incorporates additional item information to alleviate the cold-start problem and enhance the accuracy of recommendation algorithms. The proposed approach employs the BERT pre-trained model to process textual data and integrates review-based features of users and items with latent features obtained through matrix factorization. Finally, the Kolmogorov-Arnold network is utilized for feature learning in the rating prediction task. Comparative experiments on Amazon review datasets demonstrate that the proposed algorithm significantly outperforms baseline methods in terms of rating prediction accuracy and recall. By deeply mining user review texts and item descriptions, this study validates their substantial impact on improving recommendation accuracy and offers new insights for recommender system research.展开更多
在当今智能化时代,嵌入式系统开发已广泛渗透日常生活的方方面面,是电子信息类学生的必修课。其中,51单片机和STM32单片机最具代表性,网络资源丰富,常被纳入学校的课程体系。然而,在参与电子设计大赛等学科竞赛时,学生常会遇到多种复杂...在当今智能化时代,嵌入式系统开发已广泛渗透日常生活的方方面面,是电子信息类学生的必修课。其中,51单片机和STM32单片机最具代表性,网络资源丰富,常被纳入学校的课程体系。然而,在参与电子设计大赛等学科竞赛时,学生常会遇到多种复杂题型,必有一道是以TI主控为芯片的硬件设计。基于此,为帮助大学生快速掌握TI系列开发板的应用,本文以自动行驶小车为例,通过理论分析、程序对比和实验验证三部分深入剖析TI芯片与常规芯片在实际应用中的差异。In today’s intelligent era, embedded system development has been widely penetrated into all aspects of daily life, and is a required course for electronic information students. Among them, 51 microcontroller and STM32 microcontroller are the most representative, rich in network resources, and are often included in the school curriculum system. However, when participating in discipline competitions such as electronic design competitions, students often encounter a variety of complex questions, one of which must be the hardware design of TI master chip. Based on this, in order to help college students quickly grasp the application of TI series development board, this paper takes the automatic driving car as an example and deeply analyzes the differences between TI chips and conventional chips in practical application through theoretical analysis, program comparison and experimental verification.展开更多
随着深度学习技术的迅速发展,关键点检测技术在医学影像分析中的应用受到广泛关注,尤其在超声、CT和MRI等医学影像中表现出巨大的潜力。文章首先回顾了传统的关键点检测技术与基于深度学习的关键点检测技术在医学影像中的应用,重点分析...随着深度学习技术的迅速发展,关键点检测技术在医学影像分析中的应用受到广泛关注,尤其在超声、CT和MRI等医学影像中表现出巨大的潜力。文章首先回顾了传统的关键点检测技术与基于深度学习的关键点检测技术在医学影像中的应用,重点分析了卷积神经网络(CNN)、Hourglass网络和Transformer模型的特点与优势;随后讨论了关键点检测在医学影像中的实际应用,包括人体姿势估计、器官与肿瘤的分割与定位等领域的应用。此外,文章还总结了当前技术面临的挑战,如数据不足、图像噪声、跨设备泛化等问题,并提出了可能的解决方案。最后,结合深度学习的最新进展,本文展望了医学影像中关键点检测技术的未来发展趋势,旨在为医学影像分析中的关键点检测技术的研究与应用提供理论支持和发展思路。With the rapid development of deep learning technology, the application of keypoint detection technology in medical image analysis has received widespread attention, especially in medical images such as ultrasound, CT, and MRI, showing great potential. The article first reviews the application of traditional keypoint detection techniques and deep learning based keypoint detection techniques in medical imaging, with a focus on analyzing the characteristics and advantages of convolutional neural networks (CNN), Hourglass networks, and Transformer models;Subsequently, the practical applications of keypoint detection in medical imaging were discussed, including human pose estimation, segmentation and localization of organs and tumors, and other fields. In addition, the article also summarizes the challenges currently faced by technology, such as severe data shortages, image noise, cross device generalization, and proposes possible solutions. Finally, based on the latest advances in deep learning, this article looks forward to the future development trends of keypoint detection technology in medical imaging, aiming to provide theoretical support and development ideas for the research and application of keypoint detection technology in medical image analysis.展开更多
近年来,随着办公业务信息系统的广泛应用,系统中存储的涉密信息量日益增多。因此,对信息系统实施全方位、全流程、全要素的安全检测与分析显得尤为重要。通过提前识别潜在的安全隐患和风险窗口,可以有效预测可能遭遇的网络攻击路径和手...近年来,随着办公业务信息系统的广泛应用,系统中存储的涉密信息量日益增多。因此,对信息系统实施全方位、全流程、全要素的安全检测与分析显得尤为重要。通过提前识别潜在的安全隐患和风险窗口,可以有效预测可能遭遇的网络攻击路径和手段。基于此,本研究深入分析并识别网络攻击的关键环节,针对系统薄弱点进行安全措施的优化与改进,提出了一种多层次安全保密融合架构,从而实现对潜在威胁的全面防控。实验结果表明,该架构显著提升了系统的安全性能和应对复杂网络攻击的能力,通过加强安全与保密设计,进一步提升网络信息系统的整体防御能力。In recent years, with the wide application of office business information system, the amount of classified information stored in the system is increasing. Therefore, it is particularly important to implement all-round, all-process, all-element security detection and analysis of the information system. By identifying potential security threats and risk windows in advance, the path and means of possible cyber attacks can be effectively predicted. Based on this, this study analyzes and identifies the key aspects of network attacks, optimizes and improves the security measures for the weak points of the system, and proposes a multi-level security and confidentiality fusion architecture to achieve comprehensive prevention and control of potential threats. Experimental results show that the architecture significantly improves the security performance of the system and the ability to cope with complex network attacks, and further enhances the overall defense capability of the network information system by strengthening the security and confidentiality design.展开更多
针对青藏高原特殊环境下虫草检测面临的复杂高原背景、目标遮挡频繁、虫草形态细长且易与自然背景混淆等挑战,本文基于YOLOv8模型提出改进方法。首先,在可变卷积(Deformable Convolution)的基础上设计双层可变卷积(Double-layer Deforma...针对青藏高原特殊环境下虫草检测面临的复杂高原背景、目标遮挡频繁、虫草形态细长且易与自然背景混淆等挑战,本文基于YOLOv8模型提出改进方法。首先,在可变卷积(Deformable Convolution)的基础上设计双层可变卷积(Double-layer Deformable Convolution),建立双层动态卷积调整机制,利用特征偏移量自适应调整卷积核的大小和形状,提高特征饱和度,缓解下采样带来的信息失衡。其次,针对虫草因遮挡导致的漏检问题,融合空间增强注意力机制(SEAM, Spatially Enhanced Attention Module),通过深度可分离卷积和残差模块增强未遮挡部分的语义特征,优化空间通道中的权重信息,有效提升模型对遮挡环境下的信息提取和检测能力。最后,引入新的检测头FASFF-head,以自适应学习多尺度特征图的空间权重,进行空间特征融合,确保多尺度特征的协调性,且在原有检测层之上添加小目标专用检测层,使得在高密草丛环境下,虫草的检测精度得到显著提升。以上实验表明,改进模型在自建藏区虫草数据集上的mAP@0.5和mAP@0.5:0.95对比YOLOv8模型分别提升4.2%和2.9%;在Flavia Dataset公开数据集上的实验结果可以发现,YOLOv8-DSEAM 除了参数量略高于YOLOv10n,mAP@0.5比YOLOv10n提高了1.3%,mAP@0.5:0.95比YOLOv10n提高了0.8%,充分地展现了改进后的模型在高密草丛场景下的检测力和泛化力。To address the challenges of caterpillar fungus detection in the complex plateau environment of the Qinghai-Tibet Plateau, including intricate high-altitude backgrounds, frequent target occlusion, and the elongated morphology of cordyceps that easily blends with natural surroundings, this paper proposes improvements based on the YOLOv8 model. First, we design a Double-layer Deformable Convolution building upon Deformable Convolution, establishing a dual-layer dynamic convolution adjustment mechanism. This utilizes feature offsets to adaptively adjust convolution kernel size and shape, enhancing feature saturation and alleviating information imbalance caused by downsampling. Second, to tackle missed detection due to occlusion, we integrate the Spatially Enhanced Attention Module (SEAM). Through depthwise separable convolution and residual modules, this enhances semantic features of unoccluded regions and optimizes weight information in spatial channels, effectively improving information extraction and detection capabilities in occluded environments. Finally, we introduce a novel FASFF-head detection head to adaptively learn spatial weights of multi-scale feature maps for spatial feature fusion, ensuring multi-scale feature coordination. Additionally, a dedicated small-target detection layer is added above the original detection layers, significantly improving detection accuracy in dense grassland environments. Experimental results demonstrate that the improved model achieves 4.2% and 2.9% increases in mAP@0.5 and mAP@0.5:0.95 respectively compared to YOLOv8 on our self-built Tibetan Cordyceps dataset. On the public Flavia Dataset, YOLOv8-DSEAM shows superior performance: while slightly higher in parameters than YOLOv10n, it improves mAP@0.5 by 1.3% and mAP@0.5:0.95 by 0.8%, fully demonstrating the enhanced detection capability and generalization power of our model in dense vegetation scenarios.展开更多
基于热电效应的可逆性,本研究提出了一种动态可重构热电阵列系统,旨在解决电池组内部温度不均的问题,并实现快速收敛与能量收集。该系统通过动态切换每个热电模块(Thermoelectric Module, TEM)的工作模式——加热、冷却或发电,结合热点...基于热电效应的可逆性,本研究提出了一种动态可重构热电阵列系统,旨在解决电池组内部温度不均的问题,并实现快速收敛与能量收集。该系统通过动态切换每个热电模块(Thermoelectric Module, TEM)的工作模式——加热、冷却或发电,结合热点追踪技术及模糊PID (Fuzzy-PID)控制算法,实现了按需精准温控,不仅有效维持了电池工作环境的温度均匀性,还最大化利用了热差进行电力回收。实验验证表明,系统温度误差仅为1.77℃,系统超调损耗仅占整体功耗的0.11%,且在能量收集模式下能够达到319 mV的最大俘获电压。这项工作提供了高效绿色的温控解决方案。Leveraging the reversibility of the thermoelectric effect, this study introduces a dynamic reconfigurable thermoelectric array system designed to solve temperature imbalance within battery packs, enabling rapid convergence and energy harvesting. The system dynamically switches the operating mode of each thermoelectric module (TEM)—heating, cooling, or power generation —combined with hotspot tracking technology and a Fuzzy-PID control algorithm, achieving on-demand precise temperature control. This not only effectively maintains temperature uniformity within the battery operating environment but also maximizes the utilization of thermal gradients for power recovery. Experimental validation shows that the temperature error is only 1.77˚C, overshoot losses account for only 0.11% of the total power consumption, and the maximum captured voltage in the energy harvesting mode reaches 319 mV. This work presents an efficient and environmentally-friendly temperature control solution.展开更多
踝泵运动是一种临床患者进行血栓预防和康复训练的重要手段,指导患者进行正确的踝泵运动不仅可以大大降低下肢静脉血栓栓塞的发生,而且可以加快患者康复进程。目前,实际康复治疗中,由于缺乏对病人踝泵运动情况的有效监测,在对康复训练...踝泵运动是一种临床患者进行血栓预防和康复训练的重要手段,指导患者进行正确的踝泵运动不仅可以大大降低下肢静脉血栓栓塞的发生,而且可以加快患者康复进程。目前,实际康复治疗中,由于缺乏对病人踝泵运动情况的有效监测,在对康复训练过程和效果无法掌握情况下,往往无法制定出最佳康复训练计划。因此,本文设计了一种基于力和角度的踝泵运动实时监测系统,来解决以上问题。该运动监测系统采用ESP32实时采集压力传感器和姿态传感器的数据,同时利用无线方式发送到上位机,在上位机通过对数据的相应处理,得出踝泵训练过程中足底用力和踝关节角度变化情况。试验结果表明,该监测系统通过采集力与角度的实时变化,可以实现踝泵运动的有效监控。Ankle pump exercise is an important method for preventing thrombosis and promoting rehabilitation in clinical patients. Guiding patients to perform proper ankle pump exercises can significantly reduce the occurrence of lower limb venous thromboembolism and accelerate the recovery process. However, in current rehabilitation treatments, due to the lack of effective monitoring of patients’ ankle pump exercise activities, it is often difficult to develop the optimal rehabilitation plan when the training process and outcomes are not well understood. Therefore, this paper designs a real-time monitoring system for ankle pump exercise based on force and angle, aiming to address the above issues. The system utilizes an ESP32 microcontroller to collect real-time data from pressure sensors and posture sensors. The collected data is wirelessly transmitted to an upper computer, where it is processed to analyze changes in plantar force and ankle joint angles during ankle pump exercises. Experimental results show that this monitoring system effectively monitors ankle pump exercises by capturing the real-time changes in force and angle.展开更多
文摘针对水下自主航行器(AUV)在复杂水下环境中进行三维路径规划时,沙猫群算法所面临的障碍物规避能力有限、迭代效率较低以及容易陷入局部最优解等问题,本研究提出了一种将沙猫群优化算法与莱维飞行方法相结合的策略,旨在提升沙猫群算法的整体性能。基于混沌映射的均匀分布特性,改进初始种群的生成策略,有效增强了群体的多样性;此外,引入互利共生机制,并结合莱维飞行策略进行调整,显著提高算法寻找全局最优解的能力。这一改进不仅提高了算法的收敛速度,也提升了求解精度。通过静态障碍与动态洋流干扰场景的仿真测试,改进的沙猫群算法(LVSCSO)在全局收敛性上显著优于PSO、GA等六类算法:最优解偏离度降低21.4%,最差解稳定性提升33.7%,平均解精度优化19.5%。结果表明,LVSCSO可有效应对复杂水下路径规划任务(如海底勘探),具备工程部署潜力。For underwater autonomous vehicle (AUV) in complex underwater environment for 3D path planning, the sand cat group algorithm facing obstacle avoidance ability, slow convergence and easily into local optimal solution, this study puts forward a sand cat group optimization algorithm and levy flight method combining strategy, aims to improve the overall performance of the sand cat group algorithm. By initializing the initial population with the consistency of chaotic mapping, the population diversity is effectively enhanced. In addition, the mutualism mechanism and the adjustment of Levy flight strategy significantly enhance the algorithm’s ability to find the global optimal solution. This improvement not only improves the convergence speed of the algorithm, but also improves the solution accuracy. Through simulation tests in scenarios of static obstacles and dynamic current interference, the improved Sand Cat Swarm Optimization algorithm (LVSCSO) significantly outperforms six types of algorithms including PSO and GA in terms of global convergence: the deviation of the optimal solution is reduced by 21.4%, the stability of the worst solution is improved by 33.7%, and the average solution accuracy is optimized by 19.5%. The results indicate that LVSCSO can effectively address complex underwater path planning tasks (such as seabed exploration) and has potential for engineering deployment.
文摘低剂量CT技术在显著降低患者辐射剂量的同时,不可避免地引入多样化的噪声与伪影,其强度与分布特性因成像条件而异,对图像质量及临床诊断准确性构成严峻挑战。传统图像去噪方法通常基于先验知识构建数学模型,虽能有效抑制部分噪声,但其优化过程依赖人工参数调谐,存在计算复杂度高、图像细节保留不足等固有缺陷。近年来,基于深度学习的去噪方法凭借其强大的非线性特征提取与端到端优化能力,在处理复杂噪声场景时展现出显著优势。本文系统性地介绍了低剂量CT图像去噪领域的研究进展:首先剖析了传统方法的理论框架及其局限性;随后重点探讨深度学习方法的技术原理、代表性模型架构及其在医学影像中的创新应用;最后,总结当前技术面临的核心挑战,并展望未来研究方向,旨在为低剂量CT成像技术的优化与临床转化提供理论依据与技术参考。While significantly reducing patient radiation exposure, low-dose CT technology inevitably introduces diverse noise and artifacts, whose intensity and distribution characteristics vary with imaging conditions, posing a serious challenge to image quality and clinical diagnostic accuracy. Traditional image denoising methods, typically based on prior knowledge to construct mathematical models, can effectively suppress some noise. However, their optimization process relies on manual parameter tuning, exhibiting inherent limitations such as high computational complexity and insufficient preservation of image details. In recent years, deep learning-based denoising methods have demonstrated significant advantages in handling complex noise scenarios, leveraging their powerful nonlinear feature extraction and end-to-end optimization capabilities. This paper systematically introduces the research progress in the field of low-dose CT image denoising: first, it analyzes the theoretical frameworks and limitations of traditional methods;then, it focuses on the technical principles of deep learning methods, representative model architectures, and their innovative applications in medical imaging;finally, it summarizes the core challenges currently faced by the technology and outlines future research directions, aiming to provide theoretical foundations and technical references for the optimization and clinical translation of low-dose CT imaging technology.
文摘基于深度学习的推荐算法逐渐成为推荐系统领域的主流研究方向。然而,大多数现有工作仅基于单一的用户与物品交互数据,并且缺乏可解释性。本文对用户评论进行充分挖掘,并且额外引入物品信息来缓解冷启动问题并提高推荐算法的准确性。该算法利用BERT预训练模型来处理文本数据,并将用户与物品的评论特征与矩阵分解得到的潜在特征相融合,最后在评分预测任务中使用Kolmogorov-Arnold网络进行特征学习。通过本文算法与其他基线算法在亚马逊评论数据集上进行对比,该算法与基线算法相比显著提高了评分预测的精度以及准确率和召回率。本研究通过深入挖掘用户评论文本和物品描述信息,证明其在提升推荐算法准确性方面的显著效果,为推荐系统的研究提供了新的思路。Recommendation algorithms based on deep learning have emerged as a prominent research in the field of recommender systems. However, most existing approaches rely solely on user-item interaction data and lack interpretability. This article thoroughly explores user reviews and incorporates additional item information to alleviate the cold-start problem and enhance the accuracy of recommendation algorithms. The proposed approach employs the BERT pre-trained model to process textual data and integrates review-based features of users and items with latent features obtained through matrix factorization. Finally, the Kolmogorov-Arnold network is utilized for feature learning in the rating prediction task. Comparative experiments on Amazon review datasets demonstrate that the proposed algorithm significantly outperforms baseline methods in terms of rating prediction accuracy and recall. By deeply mining user review texts and item descriptions, this study validates their substantial impact on improving recommendation accuracy and offers new insights for recommender system research.
文摘在当今智能化时代,嵌入式系统开发已广泛渗透日常生活的方方面面,是电子信息类学生的必修课。其中,51单片机和STM32单片机最具代表性,网络资源丰富,常被纳入学校的课程体系。然而,在参与电子设计大赛等学科竞赛时,学生常会遇到多种复杂题型,必有一道是以TI主控为芯片的硬件设计。基于此,为帮助大学生快速掌握TI系列开发板的应用,本文以自动行驶小车为例,通过理论分析、程序对比和实验验证三部分深入剖析TI芯片与常规芯片在实际应用中的差异。In today’s intelligent era, embedded system development has been widely penetrated into all aspects of daily life, and is a required course for electronic information students. Among them, 51 microcontroller and STM32 microcontroller are the most representative, rich in network resources, and are often included in the school curriculum system. However, when participating in discipline competitions such as electronic design competitions, students often encounter a variety of complex questions, one of which must be the hardware design of TI master chip. Based on this, in order to help college students quickly grasp the application of TI series development board, this paper takes the automatic driving car as an example and deeply analyzes the differences between TI chips and conventional chips in practical application through theoretical analysis, program comparison and experimental verification.
文摘随着深度学习技术的迅速发展,关键点检测技术在医学影像分析中的应用受到广泛关注,尤其在超声、CT和MRI等医学影像中表现出巨大的潜力。文章首先回顾了传统的关键点检测技术与基于深度学习的关键点检测技术在医学影像中的应用,重点分析了卷积神经网络(CNN)、Hourglass网络和Transformer模型的特点与优势;随后讨论了关键点检测在医学影像中的实际应用,包括人体姿势估计、器官与肿瘤的分割与定位等领域的应用。此外,文章还总结了当前技术面临的挑战,如数据不足、图像噪声、跨设备泛化等问题,并提出了可能的解决方案。最后,结合深度学习的最新进展,本文展望了医学影像中关键点检测技术的未来发展趋势,旨在为医学影像分析中的关键点检测技术的研究与应用提供理论支持和发展思路。With the rapid development of deep learning technology, the application of keypoint detection technology in medical image analysis has received widespread attention, especially in medical images such as ultrasound, CT, and MRI, showing great potential. The article first reviews the application of traditional keypoint detection techniques and deep learning based keypoint detection techniques in medical imaging, with a focus on analyzing the characteristics and advantages of convolutional neural networks (CNN), Hourglass networks, and Transformer models;Subsequently, the practical applications of keypoint detection in medical imaging were discussed, including human pose estimation, segmentation and localization of organs and tumors, and other fields. In addition, the article also summarizes the challenges currently faced by technology, such as severe data shortages, image noise, cross device generalization, and proposes possible solutions. Finally, based on the latest advances in deep learning, this article looks forward to the future development trends of keypoint detection technology in medical imaging, aiming to provide theoretical support and development ideas for the research and application of keypoint detection technology in medical image analysis.
文摘近年来,随着办公业务信息系统的广泛应用,系统中存储的涉密信息量日益增多。因此,对信息系统实施全方位、全流程、全要素的安全检测与分析显得尤为重要。通过提前识别潜在的安全隐患和风险窗口,可以有效预测可能遭遇的网络攻击路径和手段。基于此,本研究深入分析并识别网络攻击的关键环节,针对系统薄弱点进行安全措施的优化与改进,提出了一种多层次安全保密融合架构,从而实现对潜在威胁的全面防控。实验结果表明,该架构显著提升了系统的安全性能和应对复杂网络攻击的能力,通过加强安全与保密设计,进一步提升网络信息系统的整体防御能力。In recent years, with the wide application of office business information system, the amount of classified information stored in the system is increasing. Therefore, it is particularly important to implement all-round, all-process, all-element security detection and analysis of the information system. By identifying potential security threats and risk windows in advance, the path and means of possible cyber attacks can be effectively predicted. Based on this, this study analyzes and identifies the key aspects of network attacks, optimizes and improves the security measures for the weak points of the system, and proposes a multi-level security and confidentiality fusion architecture to achieve comprehensive prevention and control of potential threats. Experimental results show that the architecture significantly improves the security performance of the system and the ability to cope with complex network attacks, and further enhances the overall defense capability of the network information system by strengthening the security and confidentiality design.
文摘针对青藏高原特殊环境下虫草检测面临的复杂高原背景、目标遮挡频繁、虫草形态细长且易与自然背景混淆等挑战,本文基于YOLOv8模型提出改进方法。首先,在可变卷积(Deformable Convolution)的基础上设计双层可变卷积(Double-layer Deformable Convolution),建立双层动态卷积调整机制,利用特征偏移量自适应调整卷积核的大小和形状,提高特征饱和度,缓解下采样带来的信息失衡。其次,针对虫草因遮挡导致的漏检问题,融合空间增强注意力机制(SEAM, Spatially Enhanced Attention Module),通过深度可分离卷积和残差模块增强未遮挡部分的语义特征,优化空间通道中的权重信息,有效提升模型对遮挡环境下的信息提取和检测能力。最后,引入新的检测头FASFF-head,以自适应学习多尺度特征图的空间权重,进行空间特征融合,确保多尺度特征的协调性,且在原有检测层之上添加小目标专用检测层,使得在高密草丛环境下,虫草的检测精度得到显著提升。以上实验表明,改进模型在自建藏区虫草数据集上的mAP@0.5和mAP@0.5:0.95对比YOLOv8模型分别提升4.2%和2.9%;在Flavia Dataset公开数据集上的实验结果可以发现,YOLOv8-DSEAM 除了参数量略高于YOLOv10n,mAP@0.5比YOLOv10n提高了1.3%,mAP@0.5:0.95比YOLOv10n提高了0.8%,充分地展现了改进后的模型在高密草丛场景下的检测力和泛化力。To address the challenges of caterpillar fungus detection in the complex plateau environment of the Qinghai-Tibet Plateau, including intricate high-altitude backgrounds, frequent target occlusion, and the elongated morphology of cordyceps that easily blends with natural surroundings, this paper proposes improvements based on the YOLOv8 model. First, we design a Double-layer Deformable Convolution building upon Deformable Convolution, establishing a dual-layer dynamic convolution adjustment mechanism. This utilizes feature offsets to adaptively adjust convolution kernel size and shape, enhancing feature saturation and alleviating information imbalance caused by downsampling. Second, to tackle missed detection due to occlusion, we integrate the Spatially Enhanced Attention Module (SEAM). Through depthwise separable convolution and residual modules, this enhances semantic features of unoccluded regions and optimizes weight information in spatial channels, effectively improving information extraction and detection capabilities in occluded environments. Finally, we introduce a novel FASFF-head detection head to adaptively learn spatial weights of multi-scale feature maps for spatial feature fusion, ensuring multi-scale feature coordination. Additionally, a dedicated small-target detection layer is added above the original detection layers, significantly improving detection accuracy in dense grassland environments. Experimental results demonstrate that the improved model achieves 4.2% and 2.9% increases in mAP@0.5 and mAP@0.5:0.95 respectively compared to YOLOv8 on our self-built Tibetan Cordyceps dataset. On the public Flavia Dataset, YOLOv8-DSEAM shows superior performance: while slightly higher in parameters than YOLOv10n, it improves mAP@0.5 by 1.3% and mAP@0.5:0.95 by 0.8%, fully demonstrating the enhanced detection capability and generalization power of our model in dense vegetation scenarios.
文摘基于热电效应的可逆性,本研究提出了一种动态可重构热电阵列系统,旨在解决电池组内部温度不均的问题,并实现快速收敛与能量收集。该系统通过动态切换每个热电模块(Thermoelectric Module, TEM)的工作模式——加热、冷却或发电,结合热点追踪技术及模糊PID (Fuzzy-PID)控制算法,实现了按需精准温控,不仅有效维持了电池工作环境的温度均匀性,还最大化利用了热差进行电力回收。实验验证表明,系统温度误差仅为1.77℃,系统超调损耗仅占整体功耗的0.11%,且在能量收集模式下能够达到319 mV的最大俘获电压。这项工作提供了高效绿色的温控解决方案。Leveraging the reversibility of the thermoelectric effect, this study introduces a dynamic reconfigurable thermoelectric array system designed to solve temperature imbalance within battery packs, enabling rapid convergence and energy harvesting. The system dynamically switches the operating mode of each thermoelectric module (TEM)—heating, cooling, or power generation —combined with hotspot tracking technology and a Fuzzy-PID control algorithm, achieving on-demand precise temperature control. This not only effectively maintains temperature uniformity within the battery operating environment but also maximizes the utilization of thermal gradients for power recovery. Experimental validation shows that the temperature error is only 1.77˚C, overshoot losses account for only 0.11% of the total power consumption, and the maximum captured voltage in the energy harvesting mode reaches 319 mV. This work presents an efficient and environmentally-friendly temperature control solution.
文摘踝泵运动是一种临床患者进行血栓预防和康复训练的重要手段,指导患者进行正确的踝泵运动不仅可以大大降低下肢静脉血栓栓塞的发生,而且可以加快患者康复进程。目前,实际康复治疗中,由于缺乏对病人踝泵运动情况的有效监测,在对康复训练过程和效果无法掌握情况下,往往无法制定出最佳康复训练计划。因此,本文设计了一种基于力和角度的踝泵运动实时监测系统,来解决以上问题。该运动监测系统采用ESP32实时采集压力传感器和姿态传感器的数据,同时利用无线方式发送到上位机,在上位机通过对数据的相应处理,得出踝泵训练过程中足底用力和踝关节角度变化情况。试验结果表明,该监测系统通过采集力与角度的实时变化,可以实现踝泵运动的有效监控。Ankle pump exercise is an important method for preventing thrombosis and promoting rehabilitation in clinical patients. Guiding patients to perform proper ankle pump exercises can significantly reduce the occurrence of lower limb venous thromboembolism and accelerate the recovery process. However, in current rehabilitation treatments, due to the lack of effective monitoring of patients’ ankle pump exercise activities, it is often difficult to develop the optimal rehabilitation plan when the training process and outcomes are not well understood. Therefore, this paper designs a real-time monitoring system for ankle pump exercise based on force and angle, aiming to address the above issues. The system utilizes an ESP32 microcontroller to collect real-time data from pressure sensors and posture sensors. The collected data is wirelessly transmitted to an upper computer, where it is processed to analyze changes in plantar force and ankle joint angles during ankle pump exercises. Experimental results show that this monitoring system effectively monitors ankle pump exercises by capturing the real-time changes in force and angle.
