基于深度学习的视频跟踪研究进展综述

近年来深度学习迅猛发展，颠覆了语音识别、图像分类、文本理解等领域的算法设计思路。深度学习因其具备强大的特征提取能力，在图像识别领域的成绩尤为突出。然而深度学习与视频监控领域的结合并不多，由于深度模型具有多层网络结构，算法复杂度大，训练和更新模型时比较耗时，很难满足实时性要求。回顾了深度学习的发展史，介绍了最近10年来国内外深度学习主要模型，论述了基于深度学习的目标跟踪算法，指出了各算法的优缺点，最后对当前该领域存在的问题和发展前景进行了总结和展望。     

Headspace fingerprinting approach to identify the major pathway influencing volatile patterns of vinasse-cured duck processed by high pressure,as well as its impact on physicochemical and sensory attributes

As a decisive attribute, flavour could be influenced by HP treatments through multiple physical and chemical pathways within the high pressure (HP)-assisted meat curing process. This investigation aimed to identify the major pathway influencing volatile flavour patterns of two representative vinasse-cured duck (VCD) products with HP treatments (150–300 MPa/15 min), including wet and dry types, by employing headspace fingerprinting as an untargeted approach. Results suggested that HP treatments greatly lowered moisture contents and increased Warner-Bratzler shear force and thiobarbituric acid reactive substances of the cured samples. According to multivariate models, the volatile flavour patterns of the HP-processed VCD could be clearly separated from the unprocessed samples, but the VCD pressurised at different intensities represented similar volatile fingerprinting, which was validated by e-nose analysis. The discriminant analysis (OPLS-DA) model outlined vinasse-derived ethanol, acetic acid, 3-methyl-1-butanol, 2-methyl-1-butanol, phenethyl alcohol and 2-methyl-3-octanone as the major discriminant aromas across the unpressurised and pressurised samples.     

可调偏迈步式胶带自移机尾的研究

随着掘进工作面自动化程度的提高,掘进巷道速度也在日益加快,而现有的顺槽皮带机机尾移动缓慢且移动时工作量大,需要两三人配合操作,存在一定的安全隐患。针对以上问题,研发可调偏迈步式胶带自移机尾,具有与悬挂二运皮带转载机搭接行程长,可以迈步式前进,自动调偏等功能,同时,可以降低工人的劳动强度,加快巷道掘进速度。     

视频侦查中人像智能分析应用及算法优化

人像智能分析指的是对视频或录像中的人像进行结构化和可视化分析,对目标人物进行性别、年龄、发型等特征的智能识别,这项技术在视频侦查中有极高的应用价值。人像识别早期的算法是通过人工提取特征,通过学习低级视觉特征来针对不同属性进行分类学习,这种基于传统方法的模型表现常常不尽如人意。在计算机视觉领域,通过海量图像数据学习的神经网络比传统方法有更丰富的信息量和特征可以被提取。文章尝试通过深度学习技术训练神经网络模型对行人进行检测和识别,对于衣着不同的行人进行智能识别,具有更好的鲁棒性,提升了视频人像识别的准确率,拓展了人工智能技术在身份识别领域的应用。     

除尘器新型褶式滤筒的结构优化及系列化设计

以新型褶式滤筒为研究对象,选取滤筒褶数N_2、褶夹角θ、褶高h、过滤风速v为影响因子,除尘器压力损失为响应值,基于响应曲面法分析影响因子对压力损失的影响,建立滤筒除尘器压力损失的预测模型并得到滤筒褶皱结构最优参数。以滤筒内径D、筒体高度D_1为标准尺寸,依据相关国家标准设计滤筒筒体部分;以锥体高度D_2、锥体上圆直径D_3为系列尺寸,利用尺寸变化法设计锥体部分;通过Access软件建立滤筒产品数据库。结果表明:当θ=5.2°,N_2=50,v=0.01 m/s,h=0.035 m时,除尘器压力损失最低;当D_1=1 000 mm,D=320 mm时,D_2和D_3最佳值分别为600,130 mm,故采用D_2/D_1=0.6,D_3/D=0.4的比值设计相应滤筒锥体系列。     

Visible light-activated degradation of microcystin-LR by ultrathin g-C3N4 nanosheets-based heterojunction photocatalyst

Microcystins (MCs) is a harmful toxin generated by blue-green algae in water, which has seriously threatened the ecological safety of water and human body. It is urgent to develop new catalysts and techniques for the degradation of MCs. A feasible electrostatic self-assembly method was carried out to synthesize BiVO₄/g-C₃N₄ heterojunction photocatalyst with highly efficient photocatalytic ability, where BiVO₄ nanoplates with exposed {010} facets anchored to the g-C₃N₄ ultrathin nanosheets. The morphology and microstructure of the heterojunction photocatalysts were identified by XRD, SEM, TEM, XPS, and BET. The g-C₃N₄ nanosheets have huge surface area over 200 m²/g and abundant mesoporous ranging from 2-20 nm, which provides tremendous contact area for BiVO₄ nanoplates. Meanwhile, the introduction of BiVO₄ led to red-shift of the absorption spectrum of photocatalyst, which was characterized by UV-vis diffuse reflection spectroscopy (DRS). Compared with pure BiVO₄ and g-C₃N₄, the BiVO₄/g-C₃N₄ heterojunction shows a drastically enhanced photocatalytic activity in degradation of microcystin-LR (MC-LR) in water. The MC-LR could be removed within 15 minutes under the optimal ratio of BiVO₄/g-C₃N₄. The outstanding performance of the photocatalyst is attributed to synergetic effect of interface Z-scheme heterojunction and high active facets {010} of BiVO₄ nanoplates, which provides an efficient transfer pathway to separate photoinduced carriers meanwhile endows the photocatalysts with strong redox ability.     

丙烷脱沥青装置调合前沥青的综合利用

在丙烷脱沥青装置加热炉试烧装置自产调合前沥青,通过观察油枪雾化沥青的能力、火嘴的燃烧情况以及燃烧后的烟尘状况,筛选出适合沥青燃烧的油枪,最终实现调合前沥青直接做加热炉燃料、减少重油出厂量的目的。     

析湿工况下翅片管式换热器表面粉尘沉积过程的数值模型

空调器采用翅片管式换热器作为蒸发器，在制冷工况下换热器表面发生析湿及粉尘沉积，导致性能衰减。建立湿翅片表面粉尘颗粒物沉积过程的数学模型，模拟冷凝水捕集颗粒物以及湿积灰层黏附颗粒物的过程。被冷凝水捕集的颗粒物数量等于运动轨迹与水表面轮廓会出现相交的入射颗粒物的数量；后续的入射颗粒物与湿积灰层碰撞时，部分入射颗粒物会发生沉积且部分被碰撞的已沉积湿颗粒物会发生移除，这两部分的颗粒物数量相减即为被湿积灰层黏附的颗粒物数量。模拟与实验结果的对比表明，预测的湿积灰层形状与实验照片的吻合度较好，预测的单位面积颗粒物沉积质量与91%的实验数据之间的误差在±20%之间，平均误差为11.8%。     

退火温度对冷轧DP980钢力学性能的影响

利用连续退火模拟机对DP980试验钢进行了连续退火试验,并通过扫描电镜、电子背散射衍射、透射电镜等研究了不同退火温度下(775、800、825和850 ℃),试验钢显微组织的演变规律和力学性能的变化趋势。结果表明:试验钢的屈服强度随着退火温度的升高而不断增大(从705 MPa增大到850 MPa),抗拉强度和断后伸长率则随着退火温度的升高而不断减小(抗拉强度从1150 MPa减小到1030 MPa;伸长率从8.9%减小到5.3%),这与试验钢的显微组织构成和形态分布密切相关。此外,不同退火温度下,试验钢的加工硬化率曲线均呈现单调下降的趋势。     

并联有源滤波器的改进重复控制策略研究

针对三相整流负载产生的6k±1次谐波,提出静止坐标系下的改进型6k±1重复控制策略。同时将比例控制与改进型重复控制相结合,设计改进重复控制并联比例的复合控制结构。为减小改进重复控制中延时环节不为整数的影响,采用基于Lagrange插值近似的FIR滤波器逼近分数延时特性。最后对系统进行稳定性分析和详细设计方法进行推导。通过Matlab仿真验证改进重复控制策略能有效跟踪6k±1次谐波且具有良好的补偿效果,动态响应较快。