并行生成卷积网络图像修复算法

为解决修复纹理精细、背景复杂图像中大面积不连续语义缺失时存在的边缘伪影和语义不连续的缺陷,提出一种并行生成卷积的残差连接图像修复算法。将残缺图像输入一个两列平行卷积的结构修复网络得到两个具有不同感受野大小的图像分量,通过共享解码合并两个图像分量并计算输出的L2损失优化网络。将结构修复网络的输出送入包含残差连接与注意力机制的细节修复网络,融合上下文信息,改善修复细节能力。使用全局与局部鉴别器和预训练视觉几何组网络计算损失,对修复网络进行整体判别优化,增强修复结果的整体与局部一致性。在国际公认数据库上验证提出算法的性能,实验结果表明：提出算法可以有效修复复杂背景且包含精细纹理的大面积不规则缺失区域,提升图像细节、语义和结构的真实性与完整性,其峰值信噪比和结构相似度优于经典的对比算法。     

一种推广的连续情形经济投资模型

随着序列情形费用型投资模型和折扣投资模型的解决，本文对其进行连续推广，并给出一个明确的停止规则。     

Effect of the end cap on the fragment velocity distribution of a cylindrical cased charge

The prediction of the fragment velocity distribution of a cylindrical cased charge with end caps is one of the key issues to assess the damage efficiency of the warhead. However, limited work has been con-ducted to predict the fragment velocity distributions along the axis of cylindrical cased charges with end caps. This paper presents a study of the velocity distribution of fragments caused by the explosion of a cylindrical cased charge with end caps. The fragment velocity distribution and the end cap velocity were determined by an X-ray radiography method, and the axial fragment distribution was determined by witness plates. It was found that the velocities of fragments, especially near the edge, were increased when the end caps were added, and the position of maximum velocity is closer to the non-detonation end. The fragment velocities were increased, and the fragment projection range was decreased with the increase of the thickness of the end cap. A formula for fragment velocity distributions of a cylindrical cased charge with end caps, which is based on Huang's formula, was proposed by the theoretical analysis and data fitting and validated experimentally. The results indicate that the proposed formula is accurate in predicting the fragment velocity distribution along the axis of a cylindrical cased charge with end caps detonated at one end.     

一种产生雷达多假目标的卷积调制法

讨论了多假目标干扰技术特点及应用背景,分析了用卷积调制法产生多假目标的原理,讨论了卷积器的设计方法,确定了卷积器的加权系数。仿真分析表明,卷积调制法产生多假目标方法灵活,可以实现脉内和脉间密集多假目标干扰技术,验证了方法的合理性。     

Estimation of the kinetic parameters for thermal decomposition of HNIW and its adiabatic time-to-explosion by Kooij formula

A differential/integral method to estimate the kinetic parameters (apparent activation energy E_a and pre-exponential factor A) for thermal decomposition reaction of energetic materials based on Kooij formula are applied to study the nonisothermal decomposition reaction kinetics of hexanitrohexaazaisowurtzitane (HNIW) by analyzing nonisothermal DSC curve data. The apparent activation energy (E_a) obtained by the integral isoconversional non-isothermal method based on Kooij formula is used to check the constancy and validity of apparent activation energy by the differential/integral method based on Kooij formula. The most probable mechanism function of thermal decomposition reaction of HNIW is determined by a logical choice method. The equations for calculating the critical temperatures of thermal explosion (T_b) and adiabatic time-to-explosion (t_TIad) based on Kooij formula are used to calculate the values of T_b and t_TIad to evaluate the thermal safety and heat-resistant ability of HNIW. All the original data needed for analyzing the kinetic parameters are from nonisothermal DSC curves. The results show that the kinetic model function in differential form and the values of E_a and A of decomposition reaction of HNIW are 3(1 − α)[−ln(1 − α)]^2/3, 152.73 kJ mol⁻¹ and 10^11.97 s⁻¹, respectively, and the values of self-accelerating decomposition temperature (T_SADT), T_b and t_TIad are 486.55 K, 493.11 K and 52.01 s, respectively.     

管输水力摩阻不分区计算与实验研究

长输管道输送一般处于紊流区,紊流区的水力摩阻计算一直采用分区计算的方法,涉及到的公式较多,使用起来不方便,并且分区附近存在数值跃变;由于混合摩擦区的计算通常采用近似公式,误差增加,最大甚至可达百分之十几。科尔布鲁克公式被公认为是计算混合摩擦区摩阻因数的准确公式,也可近似地用于水力光滑区和完全粗糙区的计算。文中给出了基于科尔布鲁克公式的水力摩阻计算公式,用该公式统一紊流区水力摩阻计算,并对不同油品在同一管道中流动的水力摩阻损失进行了实验研究,用实验数据分析了水力摩阻计算公式的误差。研究表明,紊流区的水力摩阻计算可不分区,均采用科尔布鲁克公式,以利于工程计算,同时可提高混合摩擦区的计算精度。     

往返搜索发现概率仿真分析与应用

往返搜索是一种最常见的搜索样式,其搜索效能与搜索宽度、发现距离、搜索速度、突破速度等因素密切相关。在建立往返搜索行动模型的基础上,通过仿真实验得到耕耘数据,进而通过数据分析和处理得到往返搜索发现概率的经验公式。应用经验公式并结合声纳探测模型,对多搜索艇分段封锁搜索案例进行了优化研究。     

对象边框标注数据的弱监督图像语义分割

针对图像语义分割应用中像素级标注数据费时昂贵的问题,主要研究以对象边框标注数据为代表的弱监督模型下的图像语义分割方法。使用基于金字塔的密集采样全卷积网络提取图像的像素级特征,并用GrabCut算法转化对弱监督数据进行数据标记,通过将图像特征和标记数据进行联合训练,构建了基于金字塔密集采样全卷积网络的对象边框标注弱监督图像语义分割模型,并在公开数据集上进行了验证。实验结果表明,所构建的弱监督模型与DET3-Proposed模型、全矩形转化模型以及Bbox-Seg模型相比,达到了更好的分割效果。     

基于卷积神经网络的卫星网络协调态势评估方法

为充分发掘利用海量卫星网络数据,提高决策效率,加强空间频轨资源获取与储备的分析手段,尤其是对地球静止轨道资源的协调获取问题,提出基于机器学习算法的卫星网络态势评估策略。通过对卫星网络协调因素进行特征分析,选择卷积神经网络(Convolution Neural Network, CNN)为目标算法模型,并建立算法模型的训练数据集及Label规则,采用分裂信息增益度量方法对数据进行降维处理,建立CNN评估模型,并进行了验证分析。结果表明,CNN模型对卫星网络协调态势评估问题测试的正确率高达80%以上,具有较高的评估效能。随着数据量的增多,CNN评估效果逐步提升,是一种在卫星网络协调态势分析、资源储备的有效评估方法。     

Gabor-CNN for object detection based on small samples

Object detection models based on convolutional neural networks (CNN) have achieved state-of-the-art performance by heavily rely on large-scale training samples. They are insufficient when used in specific applications, such as the detection of military objects, as in these instances, a large number of samples is hard to obtain. In order to solve this problem, this paper proposes the use of Gabor-CNN for object detection based on a small number of samples. First of all, a feature extraction convolution kernel library composed of multi-shape Gabor and color Gabor is constructed, and the optimal Gabor convolution kernel group is obtained by means of training and screening, which is convolved with the input image to obtain feature information of objects with strong auxiliary function. Then, the k-means clustering algorithm is adopted to construct several different sizes of anchor boxes, which improves the quality of the regional proposals. We call this regional proposal process the Gabor-assisted Region Proposal Network (Gabor-assisted RPN). Finally, the Deeply-Utilized Feature Pyramid Network (DU-FPN) method is proposed to strengthen the feature expression of objects in the image. A bottom-up and a top-down feature pyramid is constructed in ResNet-50 and feature information of objects is deeply utilized through the transverse connection and integration of features at various scales. Experimental results show that the method proposed in this paper achieves better results than the state-of-art contrast models on data sets with small samples in terms of accuracy and recall rate, and thus has a strong application prospect.