一种内容完整的视频稳定算法

设计了一种基于可靠特征集合匹配的内容完整的视频稳定算法。为了避免运动前景上的特征点参与运动估计,由经典的KLT(Kanade-Lucas-Tomasi)算法提取特征点,而后基于特征有效性判定规则对特征点集合进行有效性验证以提高特征点的可靠性。利用通过验证的特征点对全局运动进行估计,得到精确的运动参数并据此对视频图像进行运动补偿。对于运动补偿造成的无定义区,首先计算当前帧的定义区与相邻帧的光流,以此为向导腐蚀无定义区;利用拼接的方法,填充仍为无定义区的像素。实验结果表明该算法对于前景物体运动具有较好的鲁棒性并能够生成内容完整的稳定视频序列。     

一种新型微机械陀螺仪结构方案设计与仿真

为了解决微机械陀螺仪驱动和检测模态耦合的问题,提出了一种新型的双质量块双线振动结构方案。根据微陀螺的结构和工作原理,对其驱动和检测模态进行了理论分析,给出了简化的动力学方程,并利ANSYS有限元分析软件进行了数值仿真,以验证设计思想的正确性。     

基于模糊理论的非处方药的满意度评价系统

为了帮助顾客从众多对症的非处方药中选出最满意的药品,提出了开发1个基于Web的非处方药满意度评价系统。系统根据顾客提供的典型症状,采用疾病与症状相结合的选药策略,查找出所有对症的非处方药,并以药品安全性、疗效、稳定性、易用性、价格作为评价因素,根据顾客的关注角度和关注程度确定各因素权重,用模型b=W·R对对症药品进行满意度评价,按满意度高低依次推荐。经实际应用,系统推荐准确度在70%以上。     

公共卫生信息平台软件工程化开发探讨

本文讲述了在公共卫生信息平台项目的开发过程中采用软件工程化方法和技术进行平台的开发。分析了软件工程化开发现状及需求,探讨了软件工程化目标和软件工程化技术路线,并给出了软件工程化开发步骤。     

基于TerraExplorer Pro的三维城市浏览系统开发

信息技术的发展大大推动了传统的二维地理信息系统的研究成果向三维空间的拓展,三维城市浏览系统作为城市三维地理信息系统的重要组成部分,基于三维可视化和地理信息系统等技术实现城市的三维浏览.TerraExplorer Pro是Skyline软件中的一个组件,具有快速显示三维海量数据的优势.利用TerraExplorer Pro在三维显示、浏览及开发等方面的优势,探究构建城市三维浏览系统的方法,并基于TerraEzplorer Pro二次开发三维城市浏览系统,旨在为城市三维地理信息系统搭建初步显示平台.     

不同土壤湿度和雨强下径流曲线模型的改进

前期土壤湿度条件和降雨强度是影响径流曲线（SCS-CN）模型径流量预测精度的重要因素。分析不同流域下二者对SCS-CN模型性能的影响,对提高模型预测精度至关重要。基于5个半干旱半湿润和湿润流域的降雨径流资料,利用偏相关分析和K-均值聚类改进SCS-CN模型。结果表明：在重新划分前期土壤湿度条件区间后,模型预测能力大幅度提升,有效降低模型平均偏差,纳什效率系数平均提高42.8%。基于最大10 min雨强对SCS-CN模型改进后,纳什效率系数得到一定提高,且半干旱半湿润流域的提升幅度略大于湿润流域。改进的模型在研究流域都取得较好的效果,平均偏差均低于7 mm;除呈村流域非汛期外纳什效率系数均达到0.93,平均提升89%;均方根误差平均降低29.2 mm。     

The Heterointerface between Fe1/NC and Selenides Boosts Reversible Oxygen Electrocatalysis

The rational design and construction of efficient and inexpensive bifunctional oxygen electrocatalysts are highly desirable for the development of rechargeable Zn–air batteries (ZABs). Although single-atom Fe sites anchored on N-doped carbon catalysts (Fe₁/NC) ensure high oxygen reduction reaction activity, their unitary atomically dispersed active center faces difficult condition in catalyzing oxygen evolution reaction simultaneously. Herein, a composite catalyst containing heterointerface between Fe₁/NC and selenides ((Fe,Co)Se₂) is constructed. The obtained (Fe,Co)Se₂@Fe₁/NC exhibits extremely narrow potential gap of 0.616 V and remarkable stability in alkaline media, outperforming the benchmark catalysts (Pt/C+RuO₂: 0.720 V). Experimental results and density functional theory calculations reveal that heterointerface between Fe₁/NC and (Fe,Co)Se₂ accelerates the electron transfer and provides more moderate adsorption sites, which endow (Fe,Co)Se₂@Fe₁/NC with extremely high bifunctional oxygen catalytic activity. This study not only provides a superior bifunctional catalyst for ZABs, but also enriches the application of single-atom catalysts in multifunctional energy storage and conversion devices.     

Femtosecond Laser 4D Printing of Light-Driven Intelligent Micromachines

Intelligent micromachines that respond to external light stimuli have a broad range of potential applications, such as microbots, biomedicine, and adaptive optics. However, artificial light-driven intelligent micromachines with a low actuation threshold, rapid responsiveness, and designable and precise 3D transformation capability remain unachievable to date. Here, a single-material and one-step 4D printing strategy are proposed to enable the nanomanufacturing of agile and low-threshold light-driven 3D micromachines with programmable shape-morphing characteristics. The as-developed carbon nanotube-doped composite hydrogel simultaneously enhanced the light absorption, thermal conductivity, and mechanical modulus of the crosslinked network, thus significantly increasing the light sensitivity and response speed of micromachines. Moreover, the structural design and assembly of asymmetric microscale mechanical metamaterial unit cells enable the highly efficient additive nanomanufacturing of 3D shape-morphable micromachines with large dynamic modulation and spatiotemporal controllability. Using this strategy, the world's smallest artificial beating heart with programmable light-stimulus responsiveness for the cardiac cycle is successfully printed. This 4D printing method paves the way for the construction of multifunctional intelligent micromachines for bionics, drug delivery, integrated microsystems, and other fields.     

A Materials Perspective on Direct Recycling of Lithium-Ion Batteries: Principles,Challenges and Opportunities

As the dominant means of energy storage technology today, the widespread deployment of lithium-ion batteries (LIBs) would inevitably generate countless spent batteries at their end of life. From the perspectives of environmental protection and resource sustainability, recycling is a necessary strategy to manage end-of-life LIBs. Compared with traditional hydrometallurgical and pyrometallurgical recycling methods, the emerging direct recycling technology, rejuvenating spent electrode materials via a non-destructive way, has attracted rising attention due to its energy efficient processes along with increased economic return and reduced CO₂ footprint. This review investigates the state-of-the-art direct recycling technologies based on effective relithiation through solid-state, aqueous, eutectic solution and ionic liquid mediums and thoroughly discusses the underlying regeneration mechanism of each method regarding different battery chemistries. It is concluded that direct regeneration can be a more energy-efficient, cost-effective, and sustainable way to recycle spent LIBs compared with traditional approaches. Additionally, it is also identified that the direct recycling technology is still in its infancy with several fundamental and technological hurdles such as efficient separation, binder removal and electrolyte recovery. In addressing these remaining challenges, this review proposes an outlook on potential technical avenues to accelerate the development of direct recycling toward industrial applications.     

融合先验信息和特征约束的杆塔螺栓缺陷检测

目的 螺栓是输电线路中数量最多的紧固件,一旦出现缺陷就会影响电力系统的稳定运行。针对螺栓缺陷自动检测中存在的类内多样性和类间相似性挑战,提出了一种融合先验信息和特征约束的Faster R-CNN（fasterregions with convolutional neural network）模型训练方法。方法 在航拍巡检图像预处理阶段,设计了基于先验信息的感兴趣区域提取算法,能够提取被识别目标的上下文区域,从而减少模型训练阶段的数据量,帮助模型在训练阶段关注重点区域,提高其特征提取能力。在模型训练阶段,首先通过费舍尔损失约束Faster R-CNN模型的输出特征生成,使样本特征具有较小的类内距离和较大的类间间隔;然后采用K近邻算法处理样本特征得到K近邻概率,将其作为难易样本的指示以引导模型后续更加关注难样本。结果 在真实航拍巡检图像构建的螺栓数据集上进行测试,与基线模型相比,本文模型使螺栓识别的平均精度均值（mean average precision,mAP）提高了6.4%,其中正常螺栓识别的平均精度（average precision,AP）提高了0.9%,缺陷螺栓识别的平均精度提高了12%。结论 提出的融合先验信息和特征约束的输电杆塔螺栓缺陷检测方法在缺陷螺栓识别上获得了良好的效果,为实现输电线路螺栓缺陷的自动检测奠定了良好的基础。