基于改进YOLOv5s-pose的多人人体姿态估计

为了提高多人人体姿态检测的准确率,本研究采用YOLOv5s模型用于多人人体姿态检测并对模型进行改进。首先,引入坐标注意力(Coordinate Attention)模块改进骨干网络,将注意力资源分配给关键区域,降低复杂环境中的背景干扰,增强模型对多人目标的精准定位能力。其次,使用双向特征金字塔网络改进YOLOv5s的特征融合网络,增强网络的信息表达能力。实验结果表明:在多人人体姿态MS COCO2017验证集上,经改进的YOLOv5s算法的检测平均精度高达61.9%,相比原始YOLOv5s网络,平均精度提升了1.5%。由此可见,改进后的网络能更加精准、有效地检测多人人体姿态。

Multi-features Based Approach for Moving Shadow Detection

In the video-based surveillance application, moving shadows can affect the correct localization and detection of moving objects. This paper aims to present a method for shadow detection and suppression used for moving visual object detection. The major novelty of the shadow suppression is the integration of several features including photometric invariant color feature, motion edge feature, and spatial feature etc. By modifying process for false shadow detected, the averaging detection rate of moving object reaches above 90% in the test of Hall-Monitor sequence.

The Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST)

The Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST, also called the Guo Shou Jing Telescope) is a special reflecting Schmidt telescope. LAMOST’s special design allows both a large aperture (effective aperture of 3.6 m–4.9 m) and a wide field of view (FOV) (5°). It has an innovative active reflecting Schmidt configuration which continuously changes the mirror’s surface that adjusts during the observation process and combines thin deformable mirror active optics with segmented active optics. Its primary mirror (6.67m×6.05 m) and active Schmidt mirror (5.74m×4.40 m) are both segmented, and composed of 37 and 24 hexagonal sub-mirrors respectively. By using a parallel controllable fiber positioning technique, the focal surface of 1.75 m in diameter can accommodate 4000 optical fibers. Also, LAMOST has 16 spectrographs with 32 CCD cameras. LAMOST will be the telescope with the highest rate of spectral acquisition. As a national large scientific project, the LAMOST project was formally proposed in 1996, and approved by the Chinese government in 1997. The construction started in 2001, was completed in 2008 and passed the official acceptance in June 2009. The LAMOST pilot survey was started in October 2011 and the spectroscopic survey will launch in September 2012. Up to now, LAMOST has released more than 480 000 spectra of objects. LAMOST will make an important contribution to the study of the large-scale structure of the Universe, structure and evolution of the Galaxy, and cross-identification of multiwaveband properties in celestial objects.

A Comparative Study of Acute Hemodynamic Effects of Multi -site Cardiac Pacing in patients with cardiac function NYHA class ⅠtoⅡ without bundle branch block

Objective To compare the acute hemodynamic effects of five different pacing modes in patients with cardiac function NYHA class Ⅰ to Ⅱ without bundle branch block (BBB). Methods This study included 12 patients (SSS 7, Ⅲ°AVB 5) undergoing pacemaker implantation. Right ventricular apex (RVA), right ventricular outflow tract (RVOT), right ventricular bifocal (RV-Bi), left ventricular base (LVB) and bi -ventricular (Bi-Ⅴ) pacing at 60 -80 ppm were done in VVI mode prior to implantation of DDD pacemaker. The cardiac index (CI), mean pulmonary artery pressure (mPAP) and pulmonary capillary wedge pressure (PCWP) were measured with Swan - Ganz thermodilution catheter after 5 minutes of each pacing mode. Results (1) Comparing to pacing at RVA (CI: 2. 41± 0. 38 L/min per m2, PCWP: 16. 7 ±3.3 mmHg), the CI increased and the PCWP decreased significantly in pacing at RVOT(CI: 2. 63 ± 0.46, PCWP: 13. 8±2. 3), LVB(CI: 2. 78±0.52, PCWP: 14. 4±3.1), RV-Bi(CI: 2. 83±0.57, PCWP: 12. 8± 2. 5) and Bi -Ⅴ pacing (CI: 2. 94± 0.60, PCWP: 12. 7±2. 5), P < 0. 01, respectively. (2) The CI of RV-Bi and Bi-Ⅴ pacing was higher than that of RVOT and LVB pacing, the PCWP was lower, P < 0. 05, respectively. (3) There was no significant difference between RV - Bi pacing and Bi-Ⅴ pacing in CI and PCWP. Conclusion There is no significant difference between RV - Bi pacing and Bi -V pacing in the acute hemodynamic effects; however,dual - site pacing is much better than single site pacing in that aspect for patients with cardiac function NYHA class Ⅰ to Ⅱ without BBB. Among single site pacing, the RVOT and LVB pacing is better than RVA pacing in cardiac function.

Research on A Master - slave Multi - microcomputers Control System for Hollow Spindle Fancy Yarn Spinning Machine

In this paper, a successfully studied and developed master - slave muld - microcomputers control system based on PC - BUS for hollow spindle fancy yarn spinning machine, mainly Its overall scheme, software and hardware construction, is introduced. Spinning experiments show that the system achieves satisfactory result. This system can solve the diftkultles of mechatronical fusion between domestic hollow splndk fancy yarn spuming muchine and its microcomputer control technology.

On the Representation of Multi-layer Woven Structure

A quantitative method is developed to identify the structures of multi-layer woven fabric in this study. Six structural parameters are selected to represent the fabric structure. With some design criteria for the reinforcement of textile composites, the relationships among the structural parameters are derived for different binding patterns of the fabric structure. The conditions, which meet the requirements of structural cell stability,are also discussed.

基于单-多视图优化的足球球员三维姿态和体型估计

足球比赛场景的三维重建有助于观众自由切换视角,增加了互动性和沉浸感。针对足球比赛场景中的足球球员,提出一种三维姿态和体型估计方法。对球员的多视图图像使用训练好的部分注意力回归的三维人体估计(PARE)模型生成初始的三维姿态和体型估计,并使用人工标注的二维关节点作为优化目标。单-多视图优化操作利用蒙皮多人线性模型(SMPL)和正交投影的可微性,将球员的三维姿态和体型参数映射到二维关节点,计算其与人工标注之间的差异,再使用神经网络的反向传播算法更新三维姿态和体型参数,持续这些过程直到差异最小化。在自建的足球球员多视图数据集上的实验结果表明,该方法能够有效估计足球球员的三维姿态和体型,与人体网格恢复、在循环中优化SMPL、PARE等方法相比,二维关节点精度在单视图上提高了9.2%~37.5%,在多视图交叉验证中提高了34.9%~54.1%。

Computer-aided Pattern Design of Multi-bar Warp Knitted Fabrics

To meet the development trend of multi-bar warp knit-ting machine towards high-speed,advanced technologyand computer control and the requirements of variousproducts with small quantity,there are many researcheson the computer-aided pattern design of multi-barwarp knitted fabrics.In terms of the special propertiesof the computer-aided pattern design of multi-barwarp knitted fabrics,the Object Oriented Program(OOP)programming-Object Windows class Library(OWL)programming is selected.According to thecharacters of the OWL programming,various functionsare defined.Pattern design and technical parameters canbe output,which offers a great convenience for the fac-tory.

MODS： A Novel Metaheuristic of Deterministic Swapping for the Multi-Objective Optimization of Combinatorials Problems

This paper states a new metaheuristic based on Deterministic Finite Automata （DFA） for the multi - objective optimization of combinatorial problems. First, a new DFA named Multi - Objective Deterministic Finite Automata （MDFA） is defined. MDFA allows the representation of the feasible solutions space of combinatorial problems. Second, it is defined and implemented a metaheuritic based on MDFA theory. It is named Metaheuristic of Deterministic Swapping （MODS）. MODS is a local search strategy that works using a MDFA. Due to this, MODS never take into account unfeasible solutions. Hence, it is not necessary to verify the problem constraints for a new solution found. Lastly, MODS is tested using well know instances of the Bi-Objective Traveling Salesman Problem （TSP） from TSPLIB. Its results were compared with eight Ant Colony inspired algorithms and two Genetic algorithms taken from the specialized literature. The comparison was made using metrics such as Spacing, Generational Distance, Inverse Generational Distance and No-Dominated Generation Vectors. In every case, the MODS results on the metrics were always better and in some of those cases, the superiority was 100%.

基于AAM-SIFT特征描述的两级SVM人脸表情识别

采用AAM定位特征点、尺度不变特征变换(SIFT)描述特征的方式提出一种基于AAM-SIFT的表情特征提取方法。该方法用特征点周围区域梯度方向直方图描述表情特征;同时根据不同子区域对表情的贡献不同,将特征点分组并赋予不同权重,并用两级支持向量机(SVM)对融合的加权特征进行分类识别。在标准表情库和多姿态表情库上的验证结果表明,该方法能有效提高正面人脸表情的识别率,对一定偏转角度的非正面人脸表情也保持较好的鲁棒性。

G-M模型病态问题的多步解法

针对G-M模型的病态问题,分析了导致模型病态的原因和各种解决办法,重点研究了两步解法的计算过程.通过对两步解法的借鉴和引申,提出了克服模型病态性的新方法——多步解法.模拟了1个病态问题算例,应用maple数值计算软件,采用不同的方法计算该模型的参数估值,比较它们计算结果的优劣.结果表明,多步解法比两步解法计算精度更高,并且其计算步数与阈值的大小有关.

基于改进Faster R-CNN的手部位姿估计方法

基于视觉的手部位姿估计技术应用于诸多领域,具备着广泛的国际应用市场前景和巨大发展潜力。然而,手部自身存在检测目标过小、手指高自由度以及手部自遮挡等问题。通过对目前存在的难点分析,将手部位姿估计任务分为手部检测和手部关键点检测,提出基于改进的Faster R-CNN的手部位姿估计方法。首先提出基于改进的Faster R-CNN手部检测网络,将传统Faster R-CNN网络中的对ROI(regional of interest)的最大值池化,更改为ROI Align,并增加损失函数用于区分左右手。在此基础上增加了头网络分支用以训练输出MANO(hand model with articulated and non-rigid deformations)手部模型的姿态参数和形状参数,得到手部关键点三维坐标,最终得到手部的三维位姿估计结果。实验表明,手部检测结果中存在的自遮挡和尺度问题得到了解决,并且检测结果的准确性有所提高,本文手部检测算法准确率为85%,比传统Faster R-CNN算法提升13%。手部关键点提取算法在MSRA、ICVL、NYU三个数据集分别取得关键点坐标的均方误差值(key-point mean square error,KMSE)为7.50、6.32、8.50的结果。

基于优化Fast-SLAM的煤矿水泵房巡检机器人研究

针对目前煤矿水泵房无GPS环境下巡检机器人的自主定位、抗干扰能力弱、不能很好的帮机器人完成日常巡检任务等问题,研究基于Fast-SLAM的同步定位与建图方法。首先建立SLAM的概率模型,采用最小线性二乘对里程计进行标定和融合轮式里程计、IMU的方式对传感器数据进行预处理,获得比单一轮式里程计更为精确的位姿信息,减小相对定位中的非系统误差。然后再利用机器人的坐标变换关系将外部传感器获得的环境信息和机器人的运动信息转换为世界坐标系下,基于粒子滤波算法对proposal分布进行优化以获得机器人的精确位姿估计。最后通过覆盖栅格建图建立全局环境地图。经MATLAB仿真和实验结果分析表明,在经过传感器预处理和算法优化后的Gmapping算法满足机器人实时定位的需要,且具有较高的建图精度,可以有效应用于煤矿水泵房机器人自动巡检中。

视觉-激光混合式3D测量方法及自动装配研究

自动化装配作业面临基座螺纹孔位置不固定和大臂工件在搬运过程中位姿变化的问题。该文采用激光测量与视觉检测相结合的方法,实现大臂工件三维位姿测量,解决了工件搬运过程中位姿变化的问题。采用动态阈值方法提取螺纹孔连通域,通过形态学处理,提取出螺纹孔的位置特征,以装配底座为基准确定Y、Z坐标值与RX欧拉角。以基座螺纹孔轮廓作为特征模板,求取大臂与装配底座位姿的偏移量。激光传感器得到大臂工件的X方向信息,通过几何二面角求解与基准位姿的RY、RZ欧拉角偏差。最后在机器人装配系统上进行30次实验,定位姿态误差小于0.1°,位置误差小于0.2 mm,验证了提出的工件位姿测量方法的有效性。

A New Method for Pose Estimation from Line Correspondences

我们能通常在一个一般位置从三根线决定目标的姿势。三的配置非 -- 在二个点交叉的共面的线有一些特别特征，在一个一般位置的三根线没有它。这里，我们在场用这种特别的线配置决定目标姿势的一个新方法。在理论，这个方法从三线通讯充实姿势评价方法。另外，它为实际应用提供指导。而且，我们建议一个方法处理多答案现象和一个新反复的方法。模拟结果证明我们的算法快速地并且要用体力地工作。

基于poselets的特定位置人物多姿势提取

目前,对于人物识别的研究依然是一个非常具有挑战性的难题,结合多姿势来进行人物识别则是一个新的课题,因此准确提取多姿势样本是人物识别关键的一步.Poselets算法可以检测出图像中的所有人物及其相应的姿势,但是无法对特定位置的人物进行定位.因此本文提出了一种基于poselets的特定位置人物姿势提取的方法:首先根据特定位置人物头部标定框设置过滤模型,通过过滤模型对图像中由poselets算法检出的人物框进行筛选,并对筛选结果进行排序,然后结合排序得分利用二分图最大权值匹配算法对筛选结果进行匹配,找到特定位置的目标人物,提取对应的姿势.实验表明,本文算法能有效精确的检测特定位置的人物,并提取出相应的人物姿势.

Pressure Gradient Force,Saffman Lift,and Magnus Lift on the Fiber-like Particle in Fluid

Flher-like particle suspensions are common in both na-ture and industry, but there is little work reported on it.The forces acting on the fiber - like particle in fluid arestudied in this paper, and the Magnus lift, Saffman lift,pressure gradient force, and then the dynamics modelhave been received. The numerical study of the simpleshear flow past the cylinders shows that the particles ’motion is controlled by the vortex.

FP-Net:基于任意角度单幅人体图像的正面姿态估计

准确提取人体正面姿态,能够更好地辅助进行行为识别、图像生成和虚拟试衣工作.然而,侧面图像、背面图像存在人体自遮挡、关键点不可见等问题,使正面姿态提取非常困难.因此,设计并实现了基于任意角度单幅人体图像的正面姿态估计网络FP-Net(frontal pose network).首先,制作了多角度人体图像数据集,为模型设计提供数据支持;其次,为了提高模型预测结果的准确性,设计了基于Anchor pose的回归模块和基于三维姿态的特征融合模块;最后,通过FP-Net实现了任意角度人体图像的正面姿态提取.在BJUT Taichi和CMU Panoptic数据集上采用PCK评价指标进行消融实验,结果表明所提方法有效地提高了人体正面姿态估计的准确率.

Copy-Move Geometric Tampering Estimation Through Enhanced SIFT Detector Method

Digital picture forgery detection has recently become a popular and sig-nificant topic in image processing.Due to advancements in image processing and the availability of sophisticated software,picture fabrication may hide evidence and hinder the detection of such criminal cases.The practice of modifying origi-nal photographic images to generate a forged image is known as digital image forging.A section of an image is copied and pasted into another part of the same image to hide an item or duplicate particular image elements in copy-move forgery.In order to make the forgeries real and inconspicuous,geometric or post-processing techniques are frequently performed on tampered regions during the tampering process.In Copy-Move forgery detection,the high similarity between the tampered regions and the source regions has become crucial evidence.The most frequent way for detecting copy-move forgeries is to partition the images into overlapping square blocks and utilize Discrete cosine transform(DCT)com-ponents as block representations.Due to the high dimensionality of the feature space,Gaussian Radial basis function(RBF)kernel based Principal component analysis(PCA)is used to minimize the dimensionality of the feature vector repre-sentation,which improves feature matching efficiency.In this paper,we propose to use a novel enhanced Scale-invariant feature transform(SIFT)detector method called as RootSIFT,combined with the similarity measures to mark the tampered areas in the image.The proposed method outperforms existing state-of-the-art methods in terms of matching time complexity,detection reliability,and forgery location accuracy,according to the experimental results.The F1 score of the proposed method is 92.3%while the literature methods are around 90%on an average.

NUMERICAL INVERSION OF MULTIDIMENSIONAL LAPLACE TRANSFORMS USING MOMENT METHODS

This paper develops a numerical method to invert multi-dimensional Laplace transforms. By a variable transform, Laplace transforms are converted to multi-dimensional Hansdorff moment problems so that the numerical solution can be achieved. Stability estimation is also obtained. Numerical simulations show the efficiency and practicality of the method.