基于K-means/RPF的大型遮蔽空间人员定位算法

针对大型遮蔽空间惯性/地图匹配算法中粒子贫化和子粒子群迷路效应导致定位精度下降的问题,提出一种基于K-means聚类的回溯粒子滤波(RPF)人员定位算法。首先,用行人航位推算(PDR)中航向更新、步频检测及步长估计得到初始运动轨迹;然后,设计RPF算法,提高存活粒子有效性和多样性,缓解粒子贫化,提高人员定位精度;最后,通过K-means聚类算法解决子粒子群的迷路效应,修正人员轨迹出现在非可行域的现象。实验结果表明:本文算法抑制了粒子贫化和子粒子群迷路效应,人员平均定位误差相比惯性定位和标准粒子滤波降低了81.20%和51.48%。

基于K-Means++算法和改进遗传算法的维保站维修调度方法的研究

传统的电梯维保工作模式是定期保养,即针对每台电梯而言,需要每半个月去保养一次,然而电梯设备的分布和保养进度的复杂性决定了人为排定的维保路线不能够最大限度地解决维保资源。为此提出基于K-Means++算法和改进遗传算法的维保站维修调度方法,首先使用分解法的思想将问题进行拆分,然后采用K均值聚类算法将维保任务分配到合适的维保站,建立总路程最短的维保调度模型,最后利用改进遗传算法进行求解,获取最优路线规划结果,对缩减维保工作时间,提升维保工作效率,从而提升电梯使用的安全性和可靠性有一定的意义。

基于改进K-means聚类算法的分布式储能集群划分方法

随着规模化分布式电源及储能的接入,配电网的功率返送、节点过电压等问题愈加显著,对电网规划、运行监视和调度控制等造成一定影响,也不利于储能大范围发展。为此提出一种适用于规模化分布式储能的集群划分方法,基于功率节点电压灵敏度的电气距离模块度指标,对经典K-means算法进行改进,设计节点指数法、肘部法则优化初始聚类中心选择和集群数确定。以IEEE33系统算例进行验证,结果表明所提集群划分方法具有较强的电气耦合性、准确性和运算效率。

Mean Shift跟踪算法创新实验项目设计

视频跟踪算法是计算机视觉实践课程中比较受关注的实验项目。针对突变情况下传统Mean Shift跟踪算法无法实时准确跟踪的问题,设计了基于模板更新和线性预估的Mean Shift跟踪算法创新实验项目。在模板更新策略下,引入背景模板,通过将原目标模板和背景模板与设定的阈值进行比较来对干扰因素进行判定,当干扰因素判定目标受到遮挡时,引入线性预估方程进行目标位置预测,有效解决目标在遮挡情况下跟踪丢失的问题。通过对测试视频的跟踪效果和性能进行对比分析,验证了算法在突变情况下相较于传统算法具有更好的抗干扰能力。以算法创新设计为核心,通过开放性创新实验项目的选题、设计、答辩、反馈的闭环实验过程,有效提高了学生算法创新设计能力。

基于特征分箱和K-Means算法的用户行为分析方法

针对网购用户所产生的购物行为进行分析,首先通过数据处理构建客户关系管理模型(RFM模型),在此模型的基础上采用特征分箱法和K-Means聚类两种方法对用户进行细分,并对2种模型结果进行比较分析,讨论二者的差异性和具体的应用范围和意义.其中,基于特征分箱法的RFM模型将变量转化到相似的尺度上并将变量离散化,使得用户分类标签更加清晰,也可依据各类标签分类出不同类型的用户.K-Means算法通过轮廓系数评估聚类算法质量以至于选取最优K值.本文实验分析结果可为运营商提供更加可靠直观的数据,使得运营商可以根据不同用户的不同行为进行市场细分,进而进行精准营销和服务设置.

加入跳跃连接的深度嵌入K-means聚类

现有的深度聚类算法大多采用对称的自编码器来提取高维数据的低维特征,但随着自编码器训练次数的不断增加,数据的低维特征空间在一定程度上发生了扭曲,这样得到的数据低维特征空间无法反映原始数据空间中潜在的聚类结构信息.为了解决上述问题,本文提出了一种新的深度嵌入K-means算法(SDEKC).首先,在低维特征提取阶段,在对称的卷积自编码器中相对应的编码器与解码器之间以一定的权重加入两个跳跃连接,以减弱解码器对编码器的编码要求同时突出卷积自编码器的编码能力,这样可以更好地保留原始数据空间中蕴含的聚类结构信息;其次,在聚类阶段,通过一个标准正交变换矩阵将低维数据空间转换为一个新的揭示聚类结构信息的空间;最后,本文以端到端的方式采用贪婪算法迭代优化数据的低维表示及其聚类,在6个真实数据集上验证了本文提出新算法的有效性.

基于改进K-means与机器视觉的档案数据分析技术

为了提升医疗信息系统对健康档案数据的分析效率,文中采用图像采集、降噪、配准与差分等技术提取医疗图像信息,进而有效提升信息系统的数据采集效率。同时还对传统的K-means算法加以改进,并提出了一种基于类间、类内距离的聚类初始化评价指标体系(BWP),将其应用于采集到的档案数据中,以实现快速的聚类分析。将所提算法在CUDA计算平台上进行了实现,测试结果表明,该方法的聚类精度和运行效率较现有算法均有显著提升。此外,改进后K-means算法的正确聚类样本数量占比提升了4.88%,高于现有的主流指标体系,且当聚类数k的取值为16或32时,运行时间大幅降低。

Grain-size characteristics of linear dunes on the northern margin of Qarhan Salt Lake, northwestern China

In order to clarify the formation mechanism of linear dunes on the northern margin of Qarhan Salt Lake, northwestern China, we analyzed the grain-size and sorting parameters of the dune and interdune sands. The surface sands （0-30 ram） from the dune base to the crest of both flanks and interdune corridors were sampled along transects from upwind to downwind through the dune field. The results indicated that the grain-size distribution differed at different positions between and within the dunes. The frequency curve for dune sands mainly showed a bimodal distribution, while the interdune sediments showed a trimodal distribution. The grain size distribution of the linear dunes showed a finer crest pattern, i.e. the crests were composed of sands that were generally finer, better sorted than those of base sands. In addition, at the dune field scale, the dune crest sands were tending to become much finer but sorting became worse along the downwind transects. However, the grain-size parameters of sediments in the interdune corridors showed no clear pattern. The results demonstrated that the grain size and sorting parameters exhibited a systematic change not only at the individual scale but also at the dune field scale. Our results quantitatively estimate the limited role of cohesive sediments on the formation of linear dune under unidirectional wind regime. More attention should be paid to a long-term wind regime observation, internal sedimentary structures and their formation ages.

Sediment texture and grain-size implications: the Changjiang subaqueous delta

Sediment cores were collected from the subaqueous delta of the Changjiang Estuary. Sediment grain-size profiles and their fractal dimensions were analyzed, to elucidate responses to long-term sedimentary processes. In addition, the environmental sensitive populations of grain size have been extracted. The sediment cores can be divided into two parts, according to the sedimentary structures present. The upper part （0-12 cm） is interpreted as being the active layer, which is influenced frequently by changes in the short-term hydrodynamic environment. The lower part extends from a depth of 12 cm, to the bottom of the core. The pattern of fluctuation is linked to sediment grain size. Moreover, two grain-size sensitive populations can be identified. The fine sensitive population is 6.0-7.2 μm, which is a similar grain size to the suspended sediment from up-river. The coarse sensitive population varies from 40.7 to 57.5 μm, revealing complex changes. Thus, the riverine inputs from the Changjiang River may be an important source, which contributes to seasonal fluctuations of grain-size distribution, over the area. The sediments, with grain-sizes ranging from 0.9 to 20.3 μm, are characterised by self-similar in the fractal non-scale region. The fraetal dimension is eonsistant with the grain-size parameter varatioins, which could be used as a replacement index to reveal and reconstruct the sedimentary environmental evolution.

Grain-size Characteristics of Sediment in Daniugou Peatland in Changbai Mountains,Northeast China:Implications for Atmospheric Dust Deposition

The grain-size distribution characteristics and grain-size parameters of sediment in two vertical sections of Daniugou peatland in the Changbai Mountains were systematically investigated.A comparative analysis of the sediment granularity using a discriminative function with Hongyuan peat,red clay,loess-paleosol,fluvial deposit as well as lacustrine deposit was also conducted.It turns out that the vertical section of Daniugou peat ash is primarily constituted by clay and silt particles,and the content of sand is relatively small.Grain-size frequency curves generally show a single-peak modality while a bimodal pattern is detected in the upper layer.The grain-size component and peak pattern of grain-size frequency curves also illustrate that peat ash materials were transported to the peatland by long-range aeolian dust during the deposition process,while there existed short-distance dust influence in peat deposition of the upper layer.Comparisons of grain-size parameters and the discriminative Y-value of Daniugou peat ash with those of typical aeolian sediments show close similarities,suggesting the possibility that atmospheric dust transport processes were involved in the accumulation of peat again.Moreover,the variations of grain-size distribution suggest the local environmental deterioration which is just the driving force of local dust elevation.Grain-size analysis of peatland sediment is demonstrated to be one effective method to extract information about regional and global environmental evolution,and more attention should be paid to current local ecological environment and to seeking a balance between economic development and environmental protection in Northeast China.

基于K-means聚类和BP神经网络的电梯能耗实时监测方法

针对现有方法在对电梯能耗进行监测时,存在监测精度低、用时长、监测结果不理想的问题,该文提出一种基于K-means聚类算法和BP神经网络相结合的电梯能耗实时监测方法 .在经过清洗的能耗数据中提取影响建筑能耗实时监测的主要因素特征值,利用相似系数法进行相似度计算,获取相似系数.对相似电梯能耗数据进行小波分解获取高低频序列,分别采用LSSVM-GSA检测方法和均方加权处理方法对低频和高频部分进行处理,将两个结果进行重构,得到最终的实时监测结果 .仿真实验结果表明:所提方法能够获取高精度、低耗时、高稳定性的监测结果 .

基于蚁群算法的三支k-means聚类算法

在聚类分析中,三支k-means聚类算法较具有较强的处理边界不确定数据的能力,但仍然存在对初始聚类中心敏感的问题.通过将蚁群算法和三支k-means聚类算法相结合,给出了一种基于蚁群算法的三支k-means聚类算法来解决这一问题.利用蚁群算法中随机概率选择策略和信息素的正负反馈机制,动态调整权重的方法,对三支k-means聚类算法进行优化.在UCI数据集上实验证明,该方法对聚类结果的性能指标有所提高.

基于K-means与宽度学习的肺炎图像分类算法

随着人们日常生活中肺部疾病风险的增加,肺部病变筛查变得至关重要。通过CT图像快速辅助诊断肺炎可以有效遏制病情。针对现有的肺部CT图像辅助诊断方法存在数据标记量大、训练数据耗时长以及对医疗设备计算量和内存要求高等问题,提出基于K-means与宽度学习的肺炎图像分类算法。该算法引入K-means使宽度学习系统更好地提取肺部CT图像特征,缓解随机获得节点权值的性能局限,建立与典型特征学习相关的宽度学习模型,并将算法针对公开数据集进行相关实验。实验结果表明,该模型较深度学习模型的计算量大大减小,在训练速度方面有明显优势,同时也保证了较好的分类结果,极大地降低了诊断时间;在数据有限的情况下,改进后的方法与现有主流方法相比获得了更加精确的肺炎诊断结果,提出的算法更适于嵌入医学设备等资源有限的硬件系统中。

Grain-size distribution patterns of suspended sediment in response to hydrodynamics on the Dafeng intertidal flat, Jiangsu, China

Patterns of grain-size distributions of suspended sediment in relation to resuspension, settling, and tidal processes are investigated, based on in situ measurements over the Dafeng intertidal flat, on the Jiangsu coast, in the summer of 2002 and 2003. The suspended sediment here is dominated by fine and very fine silt, with a mean grain-size of 7 - 13 μm. The patterns of the grainsize distributions of suspended sediment during a tidal cycle are characterized by two types： one stable type representing insignificant spatial and temperal variations; and the other bimedal type with significant variations. The main factors influencing the grainsize distributions include resnspension, settling, suspended sediment imported into intertidal flats during the flood phase, and the grain-size distribution of seabed sediment. Resuspension increases the coarse particle content, enhances the mean grain-size of suspended sediment, and results in grain-size distributions for the suspended sediment similar to those of seabed sediment; the settling process has opposite effects on the suspended sediment. When resuspension occurs, the gain-size distributions of suspended sediment in the lower part of water column respond significantly to the current velocity. Where the influence of resuspension and settling processes is weak, the grain-size distributions of suspended sediment appear to be stable and almost identical for the various parts of intertidal flats during different measurement periods. Such distributions are referred to the background grain-size distribution, for which the mean grain-size over the Dafeng intertidal flat is around 7 μm.

Grain-size cycles in Salawusu River valley since 150 ka BP

The palaeo-mobile dune sands and fluvio-lacustrine facies with palaeosols in Milanggouwan stratigraphic section of the Salawusu River valley situated at the southeast of the Mu Us Desert experienced abundant remarkable alternative changes of coarse and fine rhythms in grain-size since 150 ka BP, and the grain-size parameters Mz,σ, Sk, Kg and SC/D also respond to the situation of multi-fluctuational alternations between peak and valley values. Simultaneity the grain-size eigenvalues F5, F16, F25, F50, F75, F84 and F95 are respondingly manifested as greatly cadent jumpiness. Hereby, the Milanggouwan section can be divided into 27 grain-size coarse and fine sedimentary cycles, which can be regarded as a real and integreted record of climate-geological process of desert vicissitude resulted from the alternative evolvement of the ancient winter and summer monsoons of East Asia since 150 ka BP.

基于K-Means聚类与熵权TOPSIS法的岩石可爆性评价研究

露天矿山的爆破块度分布,直接影响到后续的采装、运输和破碎工作。为了控制石墨矿山不同区域爆破块度分布,基于K-means无监督聚类学习法与熵权TOPSIS评价法建立了一种新的岩石可爆性评价模型,选取岩石密度、动力能量耗散率、动态抗压强度、平均应变率、脆性指数作为评价指标,通过熵权计算,发现岩石破碎程度受脆性指数影响最大,受平均应变率影响最小。将此模型应用于实际石墨矿山,可爆性分为10个等级,统计不同分级下的岩石平均破碎粒径,发现可爆性分级等级越高平均粒径越大,有明显的分级特征,验证了模型的有效性。从爆破石墨矿石岩体类型看,岩石可爆性从易到难排序为:片岩、片麻岩、变粒岩、混合岩。结合石墨矿石微观观测结果分析可知:岩性从片岩向混合岩转变,岩石内部石墨晶质呈下降趋势,石墨矿石可爆性等级也随之越来越高。岩石密度、能量耗散率、动态抗压强度之间呈线性正相关,岩石可爆性与平均应变率、脆性指数存在负相关性。研究成果为矿山矿岩可爆性评价提供了一条新思路,对露天矿山爆破块度优化具有一定的理论和实践指导意义。

基于改进K-means聚类的轨道交通基础设施分布式光伏发电典型场景生成及出力特性分析

受限于自然条件,光伏出力具有很强的随机性。为准确评估轨道交通基础设施分布式光伏发电的光伏出力特性,提出一种基于改进K-means聚类算法的轨道交通基础设施分布式光伏发电典型场景生成方法,并基于此进行光伏出力特性分析。首先,基于分布式光伏发电设施以及气象数据,利用PVsyst软件模拟光伏发电出力数据。然后,针对基本K-means聚类算法聚类参数和初始聚类中心盲目性高的问题,结合聚类有效性指标(Density based index,DBI)和层次聚类对其进行改进并利用改进K-means聚类算法生成光伏典型日出力场景。最后,基于华中地区某地轨道交通基础设施分布式光伏系统对所提方法的有效性和优越性进行验证,并通过定性和定量分析各典型场景的出力特性揭示轨道交通基础设施分布式光伏出力的规律和特点。

光伏波动平抑下改进K-means的电池储能动态分组控制策略

针对电池储能系统(battery energy storage system,BESS)进行光伏波动平抑时寿命损耗高及荷电状态(state of charge,SOC)一致性差的问题,提出了光伏波动平抑下改进K-means的BESS动态分组控制策略。首先,采用最小最大调度方法获取光伏并网指令。其次,设计了改进侏儒猫鼬优化算法(improved dwarf mongoose optimizer,IDMO),并利用它对传统K-means聚类算法进行改进,加快了聚类速度。接着,制定了电池单元动态分组原则,并根据电池单元SOC利用改进K-means将其分为3个电池组。然后,设计了基于充放电函数的电池单元SOC一致性功率分配方法,并据此提出BESS双层功率分配策略,上层确定电池组充放电顺序及指令,下层计算电池单元充放电指令。对所提策略进行仿真验证,结果表明,所设计的IDMO具有更高的寻优精度及更快的寻优速度。所提BESS平抑光伏波动策略在有效平抑波动的同时,降低了BESS运行寿命损耗并提高了电池单元SOC的均衡性。

Vertical grain-size trend of mouth bar in lacustrine fan delta:Flume experiments

Field observational previously indicated a mouth bar of a fan delta could exhibit a fining-or coarseningupward trend,which bring a new challenge to the identification of mouth bar in subsurface studies due to the lack of morphological descriptions.Previous studies have indicated that effluent behavior in rivermouth system can affect the vertical grain-size trend of mouth bar,but the drivers and magnitude of this phenomenon are not understood.We conducted flume experiments to investigate the mechanism and controlling factors of vertical grain-size trend of mouth bar.Experiment with a steeper slope of the substrate layer,greater discharge,higher sediment/water ratio,and coarser sediment induced a finingupward trend of mouth bar,because the effluent was dominated by strong inertia.Mouth bar in the experiment with a gentler slope of the substrate layer,smaller discharge,lower sediment/water ratio,and finer sediment exhibited a coarsening-upward trend dominated by the friction-dominated effluent.The relationship between the vertical grain-size trend of mouth bar and the gradients of foreset bedding in small-scale flume models and the cut-off of 15°-18°are applicable in natural systems.Identifying depositional setting to infer depositional process in river-mouth system and analyzing the plane geometry of sandbodies are two steps in the interpretation of ancient fan deltaic rock record.

Transgressive Events since the Late Pleistocene in the Yellow River Delta: Grain-size Distribution and Palynological Results

This study deals with the relationship between sea-level changes and paleoclimatic fluctuations based on the analysis of stratigraphy, grain sizes, palynology, and radiometric dating of the Yellow River delta since the Late Pleistocene. Evidence from the sedimentary record, grain sizes, and pollen provides a paleoenvironmental history of the Late Pleistocene from the boreholes of the delta. Based on a combination of grain-size analysis with lithological studies, marine deposit units contain the intervals of 13.85–16.9, 18.5–19.69, 27.9–34.8, 36.4–37.2, 48.4–51.6, and 54.1–55.9 m, and transitional facies units contain the intervals of 10.25–13.85, 16.9–18.5, 19.69–27.9, 34.8–36.4, 37.2–48.4, 51.6–54.1, and 55.9–60 m, compared with fluvial(terrestrial facies) deposit units(3.36–10.25 m). Based on pollen analysis and pollen assemblages, there were three warm-wet periods from 9.1–0.16 ka BP, 16.1–60 ka BP, and 90.1–94.6 ka BP From the top to the bottom of the borehole, the paleoclimate has an evident fluctuation: warm and moist(Holocene Optimum) —cool and dry(Younger Dryas Event)—mild semi cool—cool and dry—warm and moist. There were three warm-wet periods from 9.1–0.16 ka BP, 16.1–60 ka BP, and 90.1–94.6 ka BP, corresponding to the Holocene Optimum stage, MIS 3, and MIS 5, respectively. The warm period allowed monsoonal evergreen and broadleaved deciduous forests that corresponded to Holocene hypsithermal climatic conditions and the Late Pleistocene climatic Optimum. Three warm-wet periods occurred in marine deposit units from 9.1–0.16 ka BP, 60.1–16.1 ka BP, and 94.6–90.1 ka BP. These periods correspond to the Cangzhou transgression, Xianxian transgression, and Huanghua transgression, respectively. From 90.1–60.1 ka BP, 17.5–9.1 ka BP, and 0. 16 ka BP–1855 AD, three dry and cold phases are recognized. The phases indicate the fluvial(flood plain) sedimentary environment, corresponding to cooler and mild dry periods based on palynological results and grain-size distribution.