基于生成对抗网络的图像清晰度提升方法

视频监控、军事目标识别以及消费型摄影等众多领域对图像清晰度有很高的要求.近年来,深度神经网络在视觉和定量评估的应用研究中取得较大进展,但是其结果一般缺乏图像纹理的细节,边缘过度平滑,给人一种模糊的视觉体验.本文提出了一种基于生成对抗网络的图像清晰度提升方法.为了更好的传递图像的细节信息,采用改进的残差块和跳跃连接作为生成网络的主体架构,生成器损失函数除了对抗损失,还包括内容损失、感知损失和纹理损失.在DIV2K数据集上的实验表明,该方法在提升图像清晰度方面有较好的视觉体验和定量评估.     

深度连续卷积神经网络模型构建与性能分析

为了提升卷积神经网络特征提取能力,设计了一种基于连续卷积的深度卷积神经网络模型.该模型采用小尺度的卷积核来更细致地提取局部特征,并借助连续的两个卷积层增加模型的非线性表达能力,结合Dropout技术降低神经元之间的相互依赖,利用抑制网络过拟合对模型进行优化.人脸表情、手写数字字符和彩色图像的目标识别实验表明,在图像较为复杂时,该模型在识别的准确性和泛化性能上比手工特征提取方法及一般的2、3层卷积结构具有明显的优势.     

基于矩阵联合分块对角化的通信信号监测

在非协作通信和军事通信对抗中,接收机需要对接收到的多个通信信号进行分离,以提取出有用的信号,这可以归为通信信号卷积盲分离问题。首先构建信号模型,并将问题转化为多个矩阵的联合分块对角化。然后提出一种新的非正交矩阵联合分块对角化算法,使用最速下降法得到迭代算法,并分析了算法的可能优化策略和计算复杂度。最后仿真实验验证了新算法的有效性和可靠性,在无须计算最优步长等条件下能够获得比现有算法更快的收敛速度。     

Segmentation of the left ventricle in cardiac cine MRI using a shape-constrained snake model

Segmentation of the left ventricle (LV) is a hot topic in cardiac magnetic resonance (MR) images analysis. In this paper, we present an automatic LV myocardial boundary segmentation method using the parametric active contour model (or snake model). By convolving the gradient map of an image, a fast external force named gradient vector convolution (GVC) is presented for the snake model. A circle-based energy is incorporated into the GVC snake model to extract the endocardium. With this prior constraint, the snake contour can conquer the unexpected local minimum stemming from artifacts and papillary muscle, etc. After the endocardium is detected, the original edge map around and within the endocardium is directly set to zero. This modified edge map is used to generate a new GVC force filed, which automatically pushes the snake contour directly to the epicardium by employing the endocardium result as initialization. Meanwhile, a novel shape-similarity based energy is proposed to prevent the snake contour from being strapped in faulty edges and to preserve weak boundaries. Both qualitative and quantitative evaluations on our dataset and the publicly available database (e.g. MICCAI 2009) demonstrate the good performance of our algorithm.     

基于平面矢量场的可视化方法的研究

平面矢量场可视化是科学计算可视化的重要研究内容。矢量场以直观的图形图像显示场的运动,透过抽象数据有效洞察其内涵本质和变化规律,广泛应用于计算流体力学、航空动力学、大气物理和气象分析等领域。对平面矢量场进行可视化的方法有很多种,常见的有点图标法、流线法、线积分卷积法和拓扑分析法等。论文主要研究常见的可视化方法,分析它们的性能,并通过实际的例子,对具体的可视化结果图像进行比较。通过该研究,能够给人们在选择可视化方法时提供依据。     

Dispersion-Enabled Symmetry Switching of Photonic Angular-Momentum Coupling

Photonic spin-orbit interactions describe the interactions between spin angular momentum and orbital angular momentum of photons, which play essential roles in subwavelength optics. However, the influence of frequency dispersion on photonic angular-momentum coupling is rarely studied. Here, by elaborately designing the contribution of the geometric phase and waveguide propagation phase, the dispersion-enabled symmetry switching of photonic angular-momentum coupling is experimentally demonstrated. This notion may induce many exotic phenomena and be found in enormous applications, such as the spin-Hall effect, optical calculation, and wavelength division multiplexing systems. As a proof-of-concept demonstration, two metadevices, a multi-channel vectorial vortex beam generator and a phase-only hologram, are applied to experimentally display optical double convolution, which may offer additional degrees of freedom to accelerate computing and a miniaturization configuration for optical convolution without collimation operation. These results may provide a new opportunity for complex vector optical field manipulation and calculation, optical information coding, light-matter interaction manipulation, and optical communication.     

CNN parameter design based on fault signal analysis and its application in bearing fault diagnosis

As a representative deep learning network, Convolutional Neural Network (CNN) has been extensively used in bearing fault diagnosis and many good results have been reported. In Prognostics and Health Management (PHM) field, the CNN’s input size is usually designed as a 1D vector or 2D square matrix, and the convolution kernel size is also defined as a square shape like 3 × 3 and 5 × 5, which are directly adopted from the image recognition. Though satisfying results can be obtained, CNN with such parameter specifications is not optimal and efficient. To this end, this paper elaborated the physical characteristics of bearing acceleration signals to guide the CNN design. First, the fault period under different fault types and shaft rotation frequency were used to determine the size of CNN’s input. Next, an exponential function was involved in fitting the envelope of decaying acceleration signal during each fault period, and signal length within different decaying ratios was used to define the CNN’s kernel size. Finally, the designed CNN was validated with the Case Western Reserve University bearing dataset and Paderborn University bearing dataset. Results confirm that the physics-guided CNN (PGCNN) with rectangular input shape and rectangular convolution kernel works better than the baseline CNN with higher accuracy and smaller uncertainty. The feasibility of designing CNN parameters with physics-guided rules derived from bearing fault signal analysis has also been verified.     

融合双注意力机制的人群计数算法

针对背景复杂、遮挡、人群分布不均等人群计数常见问题,提出了一种结合联合损失的空间-通道双注意力机制卷积神经网络模型(joint loss-based space-channel dual attention network, JL-SCDANet).该网络前端进行图像粗粒度特征提取,中间加入空间注意力机制以及通道注意力机制突出图像重点区域,后端使用可加大感受野且不丢失图像分辨率的空洞卷积提取深层二维特征.此外,该模型结合联合损失函数进行训练,以增强模型的鲁棒性.为了验证模型的改进效果,在3个公共数据集(ShanghaiTech Part B、mall和UCF＿CC＿50)上分别进行了对比实验,在ShanghaiTech Part B数据集中平均绝对误差(MAE)和均方误差(MSE)分别达到了8.13和13.13;在mall数据集中MAE、MSE达到了1.78和2.28;在UCF＿CC＿50数据集中MAE、MSE分别达到了182.12和210.24,实验结果证明了该网络在提高人数统计准确率上的有效性.     

基于改进YOLOv3的智慧足球场行人检测

由于足球比赛场景中密集人群、移动小目标居多, YOLOv3算法存在检测精确度较低且模型参数量较大等问题, 使其无法部署在资源算力有限的移动设备上, 本文提出了一种基于改进YOLOv3的行人检测方法, 将Darknet-53主干特征提取网络替换为更加高效且轻量化的GhostNet网络; 同时选取了4个尺度的检测分支层并采用K-means++算法改善anchor box的聚类效果; 添加空间金字塔池化对输入图像实现相同大小的输出; 提出CIoU损失函数来计算目标定位损失值; 添加heatmap热力图可视化并在训练中使用Mosaic数据增强. 实验结果表明, YOLOv3-GhostNet在VOC融合数据集上mAP达到90.97%的同时相比YOLOv3算法提高了1.75%, 参数量减少了约81.4%且实时检测速率提高了约1.5倍, 在小型移动设备上表现出不错的检测效果.