基于时空的弱点状运动目标检测技术

如何在复杂背景下持续有效地检测目标位置,一直是研究者们需要面对的主要挑战。本文在研究红外点状移动目标特征的基础上,根据目标无纹理,无形状的特性,提出一种改进的形态学目标增强算法,并利用目标连续时空不变性检测目标。首先,建立多尺度的图像金字塔,在每层上采用改进的形态学算法快速、粗糙定位小目标。然后进一步的根据目标在时空上的位置相关性,提出基于目标运动特征分析的精确检测方法。得到精确稳定的检测结果。最后实验结果表明与经典的形态学检测算法及其他算法相比,该技术能更有效地检测弱小目标,具有更高的鲁棒性。     

In‐plane shear damage behaviours of 2D needled C/SiC composites

Understanding the in‐plane shear behaviour of composites is essential to establish the design basis for practical applications. This study aims to investigate the shear damage behaviours of 2D needled C/SiC composites by various characterization techniques. The effect of layer arrangement on shear modulus and strength was discussed via shear stress‐strain responses. The shear strain field evolution and uniformity variation were studied by digital image correlation. It shows that the uniformity of shear strain field changes with the shear load, and the shear strain field evolution consist of 5 stages. The electrical resistivity measurement results indicate that structural deformation and damage evolution caused the electrical resistivity change. Furthermore, the damage evolution has a double effect on the electrical resistivity variation. The acoustic emission monitoring shows that the shear damage evolution is a 3‐stage nonlinear process before failure. The shear damages were categorized via acoustic characteristics. Besides, the postfailure behaviours were also discussed in this study.     

卷积神经网络用于关节角度识别与姿势评估

为了快速准确地输出各种工作姿势风险评估结果，提出采用Kinect v2与卷积神经网络识别人体各关节角度，并输出标准姿势风险的评估得分。首先使用亚像素角点提取的棋盘标定算法标定Kinect两个摄像头，其次使用改进后的双边滤波对深度图像去噪，使用卷积神经网络识别人体关节二维位置，结合深度信息获取实际三维坐标并计算人体关节角度，最后输出姿势风险评估得分。通过两种实验分别验证了提出的Kinect角度识别与姿势评估的准确性，表明该方法关节角度识别与姿势风险评估的准确率均较高，是一种低成本、高可靠性的姿势评价方法，具有一定的科学意义和工程应用价值。     

基于深度学习的彩色以及近红外图像去马赛克

单传感器捕获的彩色-近红外（RGB-NIR）图像存在光谱干扰，从而导致重建出的标准彩色图像（RGB）图像与近红外（NIR）图像存在色彩失真以及细节信息模糊。针对这个问题提出一种基于深度学习的去马赛克方法，通过引入跳远连接与稠密连接解决了梯度消失和梯度弥散问题，使得网络更容易训练，并且提升了网络的拟合能力。首先，用浅层特征提取层提取了马赛克图像的像素相关性以及通道相关性等低级特征；然后，将得到的浅层特征图输入到连续多个的残差稠密块以提取专门针对去马赛克的高级语义特征；其次，为充分利用低级特征与高级特征，将多个残差稠密块提取到的特征进行组合；最后，通过全局跳远连接恢复最终的RGB-NIR图像。在深度学习框架Tensorflow上使用公共的图像与视觉表示组（IVRG）数据集、有植被的户外多光谱图像（OMSIV）数据集和森林（Forest）三个公开数据集进行实验。实验结果表明，所提方法优于基于多级自适应残差插值、基于卷积卷积和神经神经网络以及基于深度残差U型网络的主流的RGB-NIR图像去马赛克方法。     

一种图像校正的偏振光罗盘系统的研究

导航传感器在探险、应急、精确制导武器、船舶、航空器导航和定位系统中起着关键作用。针对偏振光罗盘在倾斜状态下误差较大的问题,提出了一种利用加速度计和陀螺仪对偏振光罗盘进行图像校正的方法,设计了一种基于陀螺仪校正的偏振光罗盘系统。首先利用加速度计和陀螺仪计算得到载体姿态角,对偏振光图像数据进行图像校正,采用Stokes矢量法解算大气偏振模式分布,进而提取导航特征点,最后对特征点进行拟合解算出航向角信息。并且进行了静态和动态测试实验,实验结果表明,该算法能够有效的对偏振光罗盘姿态变化引起的误差进行补偿,可以将偏振光罗盘的航向角测量误差控制在1.86°之内,平均误差为0.09°。     

图像分类中基于分类矢量量化的视觉词袋模型

特征表示是图像识别和分类的基础，视觉词袋是一种图像的特征表示方法。分析现有视觉词典构建方法的不足，提出一种新的视觉词典构建方法。首先利用梯度方差把特征矢量分为光滑类和边缘类，然后分别针对不同类别的特征矢量进行视觉词典的构建，最后根据两类视觉词典生成视觉词袋。图像分类实验表明，提出的新方法能提高分类准确率。     

Experimental investigation of far-field human cough airflows from healthy and influenza-infected subjects

Seasonal influenza epidemics have been responsible for causing increased economic expenditures and many deaths worldwide. Evidence exists to support the claim that the virus can be spread through the air, but the relative significance of airborne transmission has not been well defined. Particle image velocimetry (PIV) and hot-wire anemometry (HWA) measurements were conducted at 1 m away from the mouth of human subjects to develop a model for cough flow behavior at greater distances from the mouth than were studied previously. Biological aerosol sampling was conducted to assess the risk of exposure to airborne viruses. Throughout the investigation, 77 experiments were conducted from 58 different subjects. From these subjects, 21 presented with influenza-like illness. Of these, 12 subjects had laboratory-confirmed respiratory infections. A model was developed for the cough centerline velocity magnitude time history. The experimental results were also used to validate computational fluid dynamics (CFD) models. The peak velocity observed at the cough jet center, averaged across all trials, was 1.2 m/s, and an average jet spread angle of θ = 24° was measured, similar to that of a steady free jet. No differences were observed in the velocity or turbulence characteristics between coughs from sick, convalescent, or healthy participants.     

Sparse non-convex Lp regularization for cone-beam X-ray luminescence computed tomography

Cone-beam X-ray luminescence computed tomography (CB-XLCT) is an attractive hybrid imaging modality, and it has the potential of monitoring the metabolic processes of nanophosphors-based drugs in vivo. However, the XLCT imaging suffers from a severe ill-posed problem. In this work, a sparse nonconvex L_p (0?p?L₁ regularization. Further, an iteratively reweighted split augmented lagrangian shrinkage algorithm (IRW_SALSA-L_p) was proposed to efficiently solve the non-convex L_p (0?p?p-values (1/16, 1/8, 1/4, 3/8, 1/2, 5/8, 3/4, 7/8) in both 3D digital mouse experiments and in vivo experiments. The results demonstrate that the proposed non-convex methods outperform L₂ and L₁ regularization in accurately recovering sparse targets in CB-XLCT. And among all the non-convex p-values, our L_p(1/4?p?相似文献