An algorithm and observability research of autonomous navigation and joint attitude determination in asteroid exploration descent stage

In this paper, an autonomous relative navigation and joint attitude determination algorithm in asteroid exploration descent stage is researched based on feature point information of perpendicular asteroid surface image observed by optical navigation camera, distance vectors from spacecraft to asteroid measured by three angled installed lidars and relative velocity increment measured by accelerometer when the relative distance vector to the centroid of asteroid can not be obtained. The inertial attitude of spacecraft is determined by sun vector, star vectors and inertial angular velocity respectively measured by sun sensor, star trackers and inertial reference unit. Also, in order to obtain measurement error model transferred from sensor noise, a covariance matrix solver considering error correlation is presented via the error model of normalized vector to first order. Numerical simulation and improved observability evaluation of filtering are undertaken to discuss the results of complete sensor observation and weak observation of lidars, and verify the effectiveness of the presented relative navigation and attitude determination algorithm.     

Fractional partial differential equation denoising models for texture image

In this paper,a set of fractional partial differential equations based on fractional total variation and fractional steepest descent approach are proposed to address the problem of traditional drawbacks of PM and ROF multi-scale denoising for texture image.By extending Green,Gauss,Stokes and Euler-Lagrange formulas to fractional field,we can find that the integer formulas are just their special case of fractional ones.In order to improve the denoising capability,we proposed 4 fractional partial differential equation based multiscale denoising models,and then discussed their stabilities and convergence rate.Theoretic deduction and experimental evaluation demonstrate the stability and astringency of fractional steepest descent approach,and fractional nonlinearly multi-scale denoising capability and best value of parameters are discussed also.The experiments results prove that the ability for preserving high-frequency edge and complex texture information of the proposed denoising models are obviously superior to traditional integral based algorithms,especially for texture detail rich images.     

优化初始聚类中心选择的K-means算法

K-means算法的聚类效果与初始聚类中心的选择以及数据中的孤立点有很大关联,具有很强的不确定性。针对这个缺点,提出了一种优化初始聚类中心选择的K-means算法。该算法考虑数据集的分布情况,将样本点分为孤立点、低密度点和核心点,之后剔除孤立点与低密度点,在核心点中选取初始聚类中心,孤立点不参与聚类过程中各类样本均值的计算。按照距离最近原则将孤立点分配到相应类中完成整个算法。实验结果表明,改进的K-means算法能提高聚类的准确率,减少迭代次数,得到更好的聚类结果。     

基于人体骨骼关键点的吸烟行为检测算法

针对公共场所的监控视频中烟头目标较小并且吸烟产生的烟雾易发散,仅依靠目标检测算法检测烟头或者烟雾来判定吸烟行为存在较大难度的问题,考虑到利用骨骼关键点来进行姿态估计的算法越来越成熟,提出一种利用人体骨骼关键点和吸烟行为之间的关系来进行吸烟行为检测的算法。该算法首先利用AlphaPose和RetinaFace分别检测出人体骨骼关键点和脸部关键点信息,根据手腕到两嘴角中点和手腕到同侧眼睛的距离之比,提出一种计算人体的吸烟动作比例（SAR）是否属于吸烟动作黄金比例（GRSA）的方法以区分吸烟与非吸烟行为;再利用YOLOv4检测视频中是否存在烟头;最后结合GRSA判定和YOLOv4的结果来确定视频中存在吸烟行为的可能性高低,作出是否有吸烟行为的判定。经过笔者录制的数据集测试,结果表明所提算法可以准确检测到吸烟行为,准确率达到92%。     

改进核密度估计的空间点密度算法

针对常用的核密度估计在计算离散点密度上存在的诸多局限,提出了空间点密度算法。该算法在保持点离散属性与初始空间位置的基础上,设定分箱规则,获取离散点最近的格网点坐标;通过遍历,统计并计算每个搜索邻域内点的数量,以离散点初始坐标与点密度值为输出结果。以USGS的美国大陆地下水资源数据集展开实验研究,采用可视化输出与时间复杂度为验证指标,与核密度估计算法进行对比验证。实验结果表明,该算法提高了点密度的识别性,获取了离散点真实的密度值,可视化效果与精度方面均优于ArcGIS 10.4.1与kde2d核密度分析的结果、运算效率优于kde2d算法。     

结合密度峰值和切边权值的自训练算法

针对自训练迭代过程中错误标记样本对算法性能的影响,提出了基于密度峰值和切边权值的自训练算法。用密度聚类方法发现数据集的空间结构,选出具有代表性的未标记样本进行标签预测。用切边权值作为统计量进行假设检验,判断样本是否被正确标记,进而用正确标记样本逐步扩充有标记样本集合,直至所有未标记样本标签预测完成。新算法既充分利用了样本数据的空间结构信息,又解决了部分样本被标记错误的问题,提高了算法的分类准确率。通过在真实数据集上实验验证了新算法的有效性。     

Groundwater impact of open cut coal mine and an assessment methodology: A case study in NSW

Large scale open cut coal mining operations have significant impacts to groundwater in surrounding areas in both active and post-mining phases. The prediction of water inflows into a surface mine excavation is one of the many components involved in mine design phase. Groundwater performance also reacts to mining activities from the operational, economic and safety implications perspective. Under NSW planning legislation, as part of the comprehensive risk assessment, a groundwater impact assessment has to be conducted for a coal project to predict and mitigate the impacts in consideration of the government requirements. In this paper, the groundwater assessment modelling of mine pits was discussed in predicting of groundwater inflows and reviewing analytical and numerical approaches. A methodology of groundwater impact assessment for an open cut mine in NSW with a three-dimensional groundwater flow model Modflow Surfact demonstrated its functions in simulating the project's impacts on the groundwater regime. The key findings with mitigations are discussed and recommended in the paper to reduce impacts on groundwater and fulfil regulation requirements in NSW.     

一种基于卷积神经网络的疲劳驾驶检测方法

疲劳驾驶检测具有重要的警示作用,对检测方法的准确性和实时性均有较高要求。为此,提出了一种基于卷积神经网络的疲劳驾驶检测方法。首先,针对车内特定使用环境,对MTCNN算法进行了加速优化,在保证高准确率的同时检测速度提升高达27倍。其次,在实现人脸特征点精确定位基础上,提出了一种基于稀少特征点快速准确提取目标区域图像的ERFP（extracting images based on rare feature points）方法。再次,利用构建的眼、嘴数据集EMSD（eye and mouth state date sets）完成了眼、嘴部状态分类模型的训练。最终,利用训练得到的模型,结合相应的判定算法,实现了疲劳驾驶的检测判定。实验结果表明,该方法在实车环境下对瞌睡和哈欠行为的判定准确率均达到了96%以上,且每秒可完成约50帧图像的检测,具备良好的实时性。     

小电流接地系统两点异相接地的故障分析

根据实际运行中发生的一起导致主变差动保护动作的故障,分析了不同接地情况下对小电流接地系统运行的影响,给出了线路电流互感器不完全星形接线时,互感器中性线断线对线路保护的影响,对调度和变电运维人员处理相关异常情况具有一定的指导意义。     

基于FME的DWG高程数据到GIS数据的无损转换

高程数据主要包括等高线和高程点.本文主要研究对数字化后的1∶500无属性DWG高程数据应用FME进行定义和转换,从而实现高程数据从Auto CAD的DWG格式到Arc GIS的GIS格式的无损和高效的转换.实验结果表明,FME对高程数据的检查和转换能够更有利于各种空间分析和DEM生成.