水下运动目标实时检测算法研究

针对水下复杂环境情况,提出了一种快速实时检测水下运动目标的算法.在分析水下图像特性的基础上,对图像序列进行帧间差值运算得到差分图像,选择合适的阈值,对差分图像进行二值化和图像分割算法.结合差分图像中目标是由于待检测目标运动形成的特点,提出了衡量二值目标形状的线度指标的定义,可以描述目标的形状,从而实现快速实时检测水下运动目标.     

A new MIP model for mine equipment scheduling by minimizing maintenance cost

Mining investment has been recognized as capital intensive due mainly to the cost of large equipment. Equipment capital costs for a given operation are usually within the order of hundreds of million dollars but may reach to billion dollars for large companies operating multiple mines. Such large investments require the optimum usage of equipment in a manner that the operating costs are minimized and the utilization of equipment is maximized through optimal scheduling. This optimum usage is required to ensure that the business remains sustainable and financially stable. Most mining operations utilize trucks to haul the mined material. Maintenance is one of the major operating cost items for these fleets as it can reach approximately one hundred million dollars yearly. There is no method or application in the literature that optimizes the utilization for truck fleet over the life of mine. A new approach based on mixed integer programming (MIP) techniques is used for annually scheduling a fixed fleet of mining trucks in a given operation, over a multi-year time horizon to minimize maintenance cost. The model uses the truck age (total hours of usage), maintenance cost and required operating hours to achieve annual production targets to produce an optimum truck schedule. While this paper focuses on scheduling trucks for mining operation, concept can be used in most businesses using equipment with significant maintenance costs. A case study for a large scale gold mine showed an annual discounted (10% rate) maintenance cost saving of over $2M and more than 16% ($21M) of overall maintenance cost reduction over 10 years of mine life, compared with the spreadsheet based approach used currently at the operation.     

贝叶斯网络仿真元模型及其在装备战场损伤仿真中的应用研究

结合装备战场损伤仿真系统,研究了贝叶斯网络仿真元模型的构建方法.从条件概率角度描述了仿真模型输入参数与输出参数之间的映射关系,研究了构建贝叶斯网络仿真元模型的可行性,分析了贝叶斯网络仿真元模型的优点;研究了贝叶斯网络仿真元模型构建过程中的关键问题,包括:元模型参数的确定、原始模型参数向贝叶斯网络节点的转化、联结强度的计算、衍生元模型的构建;针对不完全信息条件下装备战场损伤快速定位问题,研究了基于K2算法的贝叶斯网络仿真元模型构建方法;构建了某型高炮的战场损伤贝叶斯网络仿真元模型.     

基于核的k-最近邻在水下目标识别中的应用*

针对水中目标特征类型多、非线性强的特点,本文将K-KNN应用于水中目标识别。该方法采用PCA对特征矩阵进行降维,利用Kernel技巧将降维后的特征映射到高维空间进行KNN分类识别,并讨论了邻近点个数K对试验结果的影响。实际试验数据验证结果表明：与传统的KNN和BP神经网络分类器相比,K-KNN分类器的综合性能更优。     

壁厚对金属圆管撕裂卷曲耗能影响的研究

对置于锥型刚性底座上内径$D$为54.4\,mm的7种不同壁厚A3无缝圆钢管,在轴 压作用下撕裂卷曲破坏进行了较为系统的实验研究. 实验过程中记录了载荷-位移曲线, 并得到试件螺旋状残余变形模态. 对实验结果的分析中认为耗散能主要由撕裂能、塑性弯曲 变形能和摩擦能3部分构成. 分析结果表明,圆管壁厚对撕裂能和塑性弯曲变形能在总能耗 中所占比例的影响较明显,对摩擦能所占比例的影响甚微.     

Non-destructive analysis of museum objects by fibre-optic Raman spectroscopy

Raman spectroscopy is a versatile technique that has frequently been applied for the investigation of art objects. By using mobile Raman instrumentation it is possible to investigate the artworks without the need for sampling. This work evaluates the use of a dedicated mobile spectrometer for the investigation of a range of museum objects in museums in Scotland, including antique Egyptian sarcophagi, a panel painting, painted surfaces on paper and textile, and the painted lid and soundboard of an early keyboard instrument. The investigations of these artefacts illustrate some analytical challenges that arise when analysing museum objects, including fluorescing varnish layers, ambient sunlight, large dimensions of artefacts and the need to handle fragile objects with care. Analysis of the musical instrument (the Mar virginals) was undertaken in the exhibition gallery, while on display, which meant that interaction with the public and health and safety issues had to be taken into account. Experimental set-up for the non-destructive Raman spectroscopic investigation of a textile banner in the National Museums of Scotland     

基于BP神经网络误差函数优化算法的武器装备承制商信用评价模型研究

信用评价是选择武器装备承制商的重要手段.以国标为基础,结合承制商具体情况确定了信用评价指标体系.分析了传统信用评价方法的不足,对经典BP神经网络的误差函数进行优化,优化后的网络模型收敛速度更快,预测精度更高.构建BP神经网络武器装备承制商信用评价模型,仿真实验表明武器装备承制商信用评价可以选用BP神经网络模型.     

水声JANUS信号的分数低阶时频谱迁移学习识别方法

非合作第三方水下标准协议信号识别在水声通信信号识别中具有重要研究意义。针对浅海水声JANUS信号的特征提取因易受脉冲噪声和多径效应等复杂水声环境影响而导致识别率低下的问题,提出一种分数低阶时频谱和ResNet18 （Residual Network 18）相结合的迁移学习识别方法。首先,选取JANUS固定前导作为识别对象,设计分数低阶傅里叶同步压缩变换（FLOFSST）,以分数低阶操作抑制脉冲噪声,以时频重排特性增强时频集中性。其次,将基于ImageNet的ResNet18预训练模型微调,迁移至JANUS信号和常见水声信号时频图集。仿真表明所提算法在信噪比为-10 dB时JANUS信号的识别率为96.15%,能够有效抑制脉冲噪声并减小多径效应影响,比传统算法识别性能好。海试中JANUS信号识别率达90.00%,证明算法识别准确率和网络的泛化性较高。     

弦线驻波实验装置改进

本文对典型弦线驻波实验中数据不够准确、驻波不稳定及实验可重复性差等现象进行了分析,明确其根本原因在于实验装置设计方案存在缺陷.提出了实验装置改进方案,一是引入拉力传感器取代狭缝刀口和悬挂重物,改变反射波的形成机理,并以弦线张力直接测量取代砝码重力间接计算;二是引入机械结构微调弦线长度以改变弦线张力和弦线上驻波波长.实验结果证明,改进后的实验装置能明显改善上述问题,该装置还可实施研究驻波规律的多个实验方案。     

基于优势特征图像融合的水下光学图像增强

针对水下光学图像颜色失真、非均匀光照、对比度低的问题,提出基于优势特征图像融合的水下光学图像增强算法.首先,提出改进的暗通道先验算法去除退化图像中的不均匀浑浊并均衡色彩;其次,对颜色校正图像分别使用基于加权分布的自适应伽玛校正算法和限制对比度自适应直方图均衡-同态滤波算法,增强颜色校正图像对比度并使其亮度均衡;最后,定义三幅融合图像即颜色校正图像、亮度均衡图像、对比度增强图像的关联权重图,通过多尺度融合算法获得融合图像.与单一预处理算法只能解决对应的退化现象相比,该算法对单幅退化图像进行多算法处理,得到三幅优势特征图像,通过不同权重的组合最大程度地将各优势特征相结合,得到的综合效果远超各单一算法优化效果,不再局限于解决颜色失真等单一问题.将本文算法与现有算法在主观评价和客观评价两方面进行实验对比,结果表明,该算法可以有效平衡水下图像的色度、饱和度及清晰度,视觉效果接近自然场景下的图像.