基于密度分类及组合特征的人数估计算法

为了克服不同人群密度及所采用特征对人数估计的影响,提出了一种基于人群密度分类及组合特征的人数统计算法,该算法包括离线特征组合选取和在线实时估计两个阶段。在离线阶段,选取密度阈值将图像样本分为高、低密度两类,然后通过实验方法选取最优的特征组合;在线估计阶段首先通过分类器将样本分为高、低密度两类,然后利用离线阶段选取的特征组合训练得到高斯模型,并分别对两类样本进行人数估计。实验结果表明,与不分高低密度相比,平均估计误差由10.6%降至8.1%;与目前主流的人数估计算法相比,该算法的平均估计误差也更小。

基于局部密度分类的人数统计算法

针对人数统计的准确性受人群密度影响较大的问题,提出一种基于局部密度分类的人数统计算法。首先,采用基于滑动窗口的子人群分割方法完成人群的分割;其次,对分割后的子人群进行高低密度分类,利用高低密度子人群在特征上的不同,通过实验方法离线地选取大小、形状、边缘、特征点和纹理5种常用特征任意组合,分别对高低密度子人群进行训练,选出适应高、低密度子人群的特征组合和支持向量回归模型;最后,采用离线阶段选择的特征和训练好的回归模型分别对高、低密度子人群进行识别。与目前主流的人数统计算法相比,该算法的平均估计误差降低了18.9%,证明了算法的有效性。

基于密度分布单类支持向量机的科技文献分类研究

在OCSVM单分类问题上,科技文献自动分类时交叉学科的分类并未得到良好的解决,且支持向量构造的超平面未考虑到非支持向量的影响,本文提出了一种基于密度分类的单类支持向量机的分类算法,将支持向量的密度分布引入目标函数。实验结果表明,该算法能够较好的将交叉学科的科技文献进行主题分类。

融合直方图高阶统计特征与GLCM特征的室内红外图像人群密度分类

公共场所的人群密度信息在公共安全、交通管理、应急减灾等方面具有重要作用,采用红外技术,可以在拍摄人群图像时避免环境光照影响。为了实现室内场景下的红外图像人群密度分类,提出一种融合灰度直方图高阶统计特征与灰度共生矩阵特征的人群密度分类方法。首先,根据红外图像的特点,分析并提取样本图像灰度直方图的高阶统计特征,随后与提取的灰度共生矩阵特征串行融合,最后作为多分类支持向量机的输入,对不同人群密度等级进行分类。实验结果表明,提出的方法对于不同密度人群图像的分类准确率可达92.13%,同时特征向量提取简洁、算法耗时短。

雨密度感知分类引导扩张网络对单张图片去雨

由于图像中的雨线条纹具有不同形状、尺寸且分布不均匀,单一神经网络学习分布不均匀的雨密度能力弱,去雨效果不显著,对此提出雨密度感知引导扩张网络对单张图片去除雨的方法。网络分为两部分:(1)雨密度感知网络对不同密度雨的图片进行分类(大雨、中雨、小雨);(2)联合雨密度感知分类信息引导扩张网络学习不同的雨密度特征细节,用于检测雨线和去雨。实验证明了该方法在合成和真实数据集上去雨的有效性。

基于密度等级分类的田间棉铃计数算法

为了解决复杂棉田环境中的多模态棉铃计数问题,提出一种基于密度等级分类的田间棉铃计数算法.首先采用密度等级分类估计器对图像中的全局上下文信息进行编码;然后利用多列结构的密度图估计器将输入图像转换为高维特征映射;最后通过特征融合神经网络,将分类信息与高维特征映射相结合,以生成高质量的密度图,进而实现对田间棉铃进行计数.此外,构建了一个包含412幅田间棉铃图像的数据集,该数据集可根据不同的环境、年份和地域条件进行划分,以进行实验和对比.实验结果表明,所提出的算法达到了更低的计数误差,其有效性和鲁棒性均优于其他对比算法.

基于密度分类的JPS+移动机器人全局路径规划算法

针对传统全局路径规划中扩展节点多、寻路时间长等问题,提出一种基于JPS+(jump point search plus)算法的全局路径规划算法,旨在提高机器人在复杂环境的智能性、高效性的要求。首先引入了一种基于密度的判断障碍物角点规则,实现对于主要跳点的识别数目,减少搜索路径过程中的可扩展节点,同时在路径求解过程中对目标跳点的判定规则进行了修改,最终实现了减少计算量、缩短计算时长的目标。为验证所提改进型JPS+算法的有效性,将A、JPS+算法在不同类型地图中与改进型JPS+算法进行了比较。仿真结果表明,改进型JPS+算法与A算法相比,在路径长度、寻路时间和扩展节点数量上都有明显改进;在生成相同路径的基础上,与传统JPS+算法相比,在障碍物占比33.25%的地图中搜索时间降低了7.58%,节点扩展数量减少了9.38%,能够满足移动机器人快速全局路径规划的要求。

视频前景多特征融合的人群密度分析

视频GIS技术可以融合视频处理技术与GIS分析技术实现视频场景的公共安全监控,在公共场所视频监控应用领域,人群密度和数量能够反映地理空间场景的人群拥挤程度,具有一定的公共安全风险控制作用。为满足公共安全领域利用视频GIS技术对视频人群监测及分析的需求,提出了一种视频前景多特征融合的人群密度及数量估计方法。该方法融合前景图像块的边缘纹理特征,通过SVM区分场景人群密度,然后以归一化前景面积、轮廓和边缘纹理特征作为回归因子,根据人群密度差异分类构建回归模型,进而估算场景人群数量。实验结果表明,此方法能够高效估计视频场景的人群密度和数量,准确表征该场景人群分布状态。

基于纹理的车辆阴影消除新算法

基于对智能交通系统(ITS,Intelligent Transport Systems)中视频检测的研究和分析,特别针对其中关键步骤之一的阴影消除展开深入探讨,分析了阴影产生的原理和特点,阐述了现有的阴影去除算法,在现有算法的基础上提出了一种基于区域聚类的阴影消除算法。实践证明,该方法能够较好的去除运动车辆的阴影,保留较完整的车辆目标信息,为准确提取车辆目标奠定了基础。

一种基于纹理的车辆阴影消除新算法

基于对智能交通系统(ITS,Intelligent Transport Systems)中视频检测的研究和分析,特别针对其中关键步骤之一的阴影消除展开深入探讨,分析了阴影产生的原理和特点,阐述了现有的阴影去除算法,在现有算法的基础上提出了一种基于区域聚类的阴影消除算法。实践证明,该方法能够较好的去除运动车辆的阴影,保留较完整的车辆目标信息,为准确提取车辆目标奠定了基础。

基于秩次集对分析理论的空间负荷预测方法

利用秩次集对分析理论在处理系统不确定性方面的优势,提出一种新的空间负荷预测方法。首先,在电力地理信息系统中,根据待预测区域内各10 kV馈线供电范围生成Ⅰ类元胞,将Ⅰ类元胞的历史负荷数据分别按不同的集合容量生成多个历史数据集合和1个目标数据集合;其次,对各历史数据集合进行秩次变换得到相应的秩次集合,并分别将其与目标数据秩次集合构成集对;然后,寻找与目标数据集合相似的历史数据集合,选取相对误差最小的集合容量对应的预测值作为各Ⅰ类元胞负荷预测值;最后,以等大小网格生成Ⅱ类元胞,根据Ⅰ类元胞负荷预测值结合用地信息求出各Ⅱ类元胞的负荷预测值,从而得到网格化后的空间负荷预测结果。工程实例验证了所提方法的实用性和有效性。

浅析SLF中各影响因素对预测精度的重要程度

空间电力负荷发展趋势受多种变量与影响因素的制约,多种因素共同作用的结果决定了该系统的发展态势。同时,每一个变量的发展变化都不是孤立的,一个变量要受到其他变量的影响,也影响着其他变量。影响负荷预测的变量非常多,其中多数可以借助历史年鉴数据获取,但小区的空间信息需借助地理信息系统(GIS)获取。本文我们将各变量与分类负荷密度的相关程度作为各影响因素对预测精度的重要程度的评价指标。准确把握各种影响变量或因素,尤其是主要影响因素及其权重、发展变化规律,更有助于提高SLF的准确性。该类研究将对城市电网建设及其经济和安全的运行的具有深远影响。

Classification and Distribution of Common Aroma Components in Tobacco Leaf Based on Their Physical and Chemical Properties

[Objective] This study was aimed to find a new method for the classification of common aroma components in tobacco leaves. [Method] Sixty-four common aroma components in tobacco leaves were classified by cluster analysis based on their relative molecular weight. The contents and distribution of aroma components in another 71 C3F and 64 B2F tobacco leaf samples were analyzed by using the new method. [Result] The 64 common aroma components were divided into three categories trough the cluster analysis based on their molecular weight. CategoryⅠ consisted of 12 aroma components which had high molecular weight（281.308±21.536on average） and high boiling point（371.311±29.904 ℃ on average）. Category Ⅱ included 27 components which had low molecular weight（103.722 ± 13.115 on average） and low boiling point（176.132±42.342 ℃ on average）. Category Ⅲ included 25components which had middle molecular weight（175.393 ± 24.906 on average） and middle boiling point（250.562±45.431 ℃ on average）. The content of high-molecularweight aroma components in middle leaves（547.344±224.391 μg/g） was much higher than that in upper leaves（477.549±182.066 μg/g）. The content of low-molecularweight aroma component in middle leaves（17.468±3.459 μg/g） was also significantly higher than that in upper leaves（15.936±3.456 μg/g）. The content of middle-molecular-weight aroma component in middle leaves（44.931 ±8.953 μg/g） was extremely significantly higher than that in upper leaves（37.997±6.042 μg/g）. [Conclusion] This study proposed a new way to classify the aroma components in flue-cured tobacco leaves using the relative molecular weight as the index, which will provide theoretical reference for developing special tobacco leaves.

交通事故的时空热点分析

针对目前交通事故分析中时空维度分离的不足,以H市2013-2015年的交通事故数据为研究对象,根据自组织神经网络、平行坐标系、时空颜色矩阵和时空网络核密度估计、热点分析法,分别从行政区划以及道路网络上进行交通事故的时空特征分析,从宏观和微观角度揭示交通事故的时空热点区域。结果表明,整体来说,在空间上,H市交通事故热点区域4个街道高于8个镇,呈明显的“两带一中心”分布,即硖许线与101省道形成的带状区域和市政府所在的行政中心区域;时间上,早、晚高峰时段最为严重,夜晚较为轻微,呈明显的区域特征分布,即工业办公区热点出现在早高峰时段,生活住宅区热点出现在晚高峰时段,商业消费区热点出现在夜晚时段。

A New Extension Theory-based Production Operation Method in Industrial Process

To explore the problems of dynamic change in production demand and operating contradiction in production process, a new extension theory-based production operation method is proposed. The core is the demand requisition, contradiction resolution and operation classification. For the demand requisition, the deep and comprehensive demand elements are collected by the conjugating analysis. For the contradiction resolution, the conflict between the demand and operating elements are solved by the extension reasoning, extension transformation and consistency judgment. For the operating classification, the operating importance among the operating elements is calculated by the extension clustering so as to guide the production operation and ensure the production safety. Through the actual application in the cascade reaction process of high-density polyethylene (HDPE) of a chemical plant, cases study and comparison show that the proposed extension theory-based production operation method is significantly better than the traditional experience-based operation method in actual production process, which exploits a new way to the research on the production operating methods for industrial process.

A NON-PARAMETER BAYESIAN CLASSIFIER FOR FACE RECOGNITION

A non-parameter Bayesian classifier based on Kernel Density Estimation (KDE)is presented for face recognition, which can be regarded as a weighted Nearest Neighbor (NN)classifier in formation. The class conditional density is estimated by KDE and the bandwidthof the kernel function is estimated by Expectation Maximum (EM) algorithm. Two subspaceanalysis methods-linear Principal Component Analysis (PCA) and Kernel-based PCA (KPCA)are respectively used to extract features, and the proposed method is compared with ProbabilisticReasoning Models (PRM), Nearest Center (NC) and NN classifiers which are widely used in facerecognition systems. The experiments are performed on two benchmarks and the experimentalresults show that the KDE outperforms PRM, NC and NN classifiers.

模糊贴近度理论在空间负荷预测中的应用

电力系统负荷预测是配电网规划的重要依据,单纯的负荷总量预测并不能反映规划区未来负荷增长在地理上的分布,空间负荷预测的提出正好弥补了这一点。基于分类负荷密度法的空间负荷预测难点在于分类负荷密度的确定,文章引入模糊贴近度理论,使分类负荷密度的选取更加科学化。

Combined Density-based and Constraint-based Algorithm for Clustering

We propose a new clustering algorithm that assists the researchers to quickly and accurately analyze data. We call this algorithm Combined Density-based and Constraint-based Algorithm (CDC). CDC consists of two phases. In the first phase, CDC employs the idea of density-based clustering algorithm to split the original data into a number of fragmented clusters. At the same time, CDC cuts off the noises and outliers. In the second phase, CDC employs the concept of K-means clustering algorithm to select a greater cluster to be the center. Then, the greater cluster merges some smaller clusters which satisfy some constraint rules. Due to the merged clusters around the center cluster, the clustering results show high accuracy. Moreover, CDC reduces the calculations and speeds up the clustering process. In this paper, the accuracy of CDC is evaluated and compared with those of K-means, hierarchical clustering, and the genetic clustering algorithm (GCA) proposed in 2004. Experimental results show that CDC has better performance.

K-means Find Density Peaks in Molecular Conformation Clustering

Performing cluster analysis on molecular conformation is an important way to find the representative conformation in the molecular dynamics trajectories.Usually,it is a critical step for interpreting complex conformational changes or interaction mechanisms.As one of the density-based clustering algorithms,find density peaks(FDP)is an accurate and reasonable candidate for the molecular conformation clustering.However,facing the rapidly increasing simulation length due to the increase in computing power,the low computing efficiency of FDP limits its application potential.Here we propose a marginal extension to FDP named K-means find density peaks(KFDP)to solve the mass source consuming problem.In KFDP,the points are initially clustered by a high efficiency clustering algorithm,such as K-means.Cluster centers are defined as typical points with a weight which represents the cluster size.Then,the weighted typical points are clustered again by FDP,and then are refined as core,boundary,and redefined halo points.In this way,KFDP has comparable accuracy as FDP but its computational complexity is reduced from O(n^(2))to O(n).We apply and test our KFDP method to the trajectory data of multiple small proteins in terms of torsion angle,secondary structure or contact map.The comparing results with K-means and density-based spatial clustering of applications with noise show the validation of the proposed KFDP.

Automated Building Block Extraction and Building Density Classification Using Aerial Imagery and LiDAR Data

This paper examines the utility of high-resolution airborne RGB orthophotos and LiDAR data for mapping residential land uses within the spatial limits of suburb of Athens, Greece. Modem remote sensors deliver ample information from the AOI （area of interest） for the estimation of 2D indicators or with the inclusion of elevation data 3D indicators for the classification of urban land. In this research, two of these indicators, BCR （building coverage ratio） and FAR （floor area ratio） are automatically evaluated. In the pre-processing step, the low resolution elevation data are fused with the high resolution optical data through a mean-shift based discontinuity preserving smoothing algorithm. The outcome is an nDSM （normalized digital surface model） comprised of upsampled elevation data with considerable improvement regarding region filling and ＂straightness＂ of elevation discontinuities. Following this step, a MFNN （multilayer feedforward neural network） is used to classify all pixels of the AOI into building or non-building categories. The information derived from the BCR and FAR building indicators, adapted to landscape characteristics of the test area is used to propose two new indices and an automatic post-classification based on the density of buildings.