基于粒子群三次样条优化的图像边缘描述算法

针对手工选点生成的三次样条曲线无法准确地描述图像边缘的问题,以粒子群三次参数样条曲线优化算法为基础对现有的图像边缘算法进行了改进。在图像边缘上取若干点作为样条曲线的型值点,利用粒子群算法对这些型值点进行优化,使得到的样条曲线对图像边缘有很好的贴合。实验表明,该算法不仅能够快速拟合样条曲线,而且优化后的样条曲线形态能较好地贴合图像边缘轮廓。     

Zernike矩结合Sigmoid曲线拟合边缘检测方法

为了改进Zernike矩边缘检测方法对非直线边缘检测精度低的弱点,提出了一种改进的基于Zernike矩方法和Sigmoid拟合法的边缘检测方法。用Zernike矩边缘检测方法检测边缘,然后用Sigmoid拟合方法检测边缘,根据边缘类型对两种方法检测的边缘结果进行加权求和,所得结果为边缘位置。分别利用理想图像和实际工业CT图像进行了实验验证,实验结果表明对非直线边缘的检测,该方法检测精度相对于Zernike矩方法提高了一个数量级。     

一种无线传感器网络链式传输分簇路由协议

由于周围环境对无线传感器网络(WSNs)的影响,在布设到特殊环境下时会产生信号的衰减与损耗,导致通信不畅。针对此问题,提出一种以LEACH路由协议为基础适应特殊环境(长直空间)的新型路由算法。本算法采用链式传输,即从内部逐一将信号传输给距空间最外端且距基站位置较近的簇头,克服了内部节点死亡过快的问题。同时簇内采用链式传输并且改进簇头阈值与成簇半径,减小了能量消耗,提高了稳定性,克服了LEACH算法的不足。     

无线传感器网络分簇拓扑控制算法

通过对经典的分簇算法HEED和EEUC进行研究与分析,对它们不足之处进行了改进,提出了一种新的基于双簇首节能的无线传感器网络分簇拓扑控制算法,即DCHEB算法。该算法提出了一种新的簇划分方案,通过此方案可以对无线传感器网络进行合理分簇,使得簇首节点位于合适的位置上,平均了各个簇的节点个数,可以避免簇内的边缘节点过早死亡。最后通过理论分析和仿真工具验证了该算法对减少无线传感器网络的能量消耗和延长其生存时间有很好的作用。     

一种具有强实时性、强鲁棒性的图像匹配算法

针对描述符BRIEF对图像旋转敏感的问题,提出一种改进的描述符RIBRIEF,该描述符具有识别能力强、提取速度快、占用空间小及抗干扰能力强等优点,并具有旋转不变性.经分析,图像匹配算法的实时性较大程度上由特征点数量、匹配点搜索次数及描述符相似度计算复杂度决定,因此提出通过描述符索引与描述符聚类相结合、基于FAST稳定特征点提取和逻辑运算计算相似度等方法提高算法的整体实时性.实验结果表明,与描述符BRIEF及SURF相比较,基于描述符RIBRIEF的图像匹配算法在鲁棒性及实时性方面均具有明显优势.     

对边缘不突出的显微镜影像进行边缘特征提取与Hough直线检测

文章主要介绍如何对边缘不突出的显微镜进行边缘特征提取与Hough直线检测,其中涉及到数字图像处理中的边缘检测、图像二值化及Hough直线检测算法。     

基于多GPU集群的编程框架

现如今,GPU作为一种低功耗高性能图形处理器单元,被广泛应用于高度并行化的应用程序中。其线程和内存的层次结构在诸多成功的多线程应用和科学研究中表现出巨大的优势。为了简化多GPU集群的编程模式以及更好地利用GPU的计算性能,设计并实现了一个新的基于多GPU的MapReduce并行编程框架。使用了并行虚拟文件系统（PVFS）来存储数据,考虑了动态的负载平衡和GPU相关的权重要素以达到优化系统的效率、透明性以及系统的可伸缩性的目的。在文中,将演示使用该编程模式解决地质应用的一个典型的偏移应用-叠前时间偏移（PKTM）,并给出实验结果。     

基于云端协同计算架构的边缘端I/O密集型虚拟机资源分配方案

大型制造业生产车间业务流程复杂,传统固定资源配置方式将产生负载不均衡的问题。针对此问题,设计并实现了一种基于云端协同架构的边缘端I/O密集型虚拟机资源的分配算法,通过定义计算节点中每个I/O密集型虚拟机三个维度的信息计算得出I/O密集型虚拟机的优先级,并按最大优先级队列实时统一分配硬件资源。实验结果表明,该算法在应对工业机械设备高响应和高通量的要求上有较为明显的提升,为实际生产起到了优化资源配置的作用。     

Numerical study of coal particle cluster combustion under quiescent conditions

Behavior of ignition and combustion of coal particle cluster under a quiescent condition was numerically simulated by solving balance equations of mass and enthalpy with combustion kinetic models of volatiles and char. Two-flame structure, one flame penetrating into the cluster and the other moving out of the cluster, was predicted during the combustion of coal particle cluster. Effects of radiative heat transfer, group number, ambient temperature, coal particle size, and oxygen concentration on ignition and combustion of coal particle clusters were also analyzed. Simulations indicated that the gas volume fraction of coal particle cluster increases with time after devolatilization. Gas velocity passing through the cluster surface varied significantly at volatile liberation. The ignition time delay was reduced with the increase of ambient temperature. The cluster devolatilization rate and char burning rate increased while the ignition time delay decreased with the increase of ambient oxygen concentration.     

Selectivities in methanol oxidation over silica supported molybdena

Selectivities in methanol oxidation over silica supported molybdenum oxide catalysts were investigated in relation with the phase distribution. The supported catalysts were prepared by impregnation with ammonium heptamolybdate. In addition to crystalline MoO₃, Mo containing cluster species of 1–2 nm size were observed by STEM even from a used catalyst with 13% catalyst loading. The percentage of Mo present as crystalline MoO₃ increases with the catalyst loading. An ESCA study indicates that part of surface Mo in the supported catalysts is reduced to Mo⁵⁺. The dimethyl ether selectivity increases with the catalyst loading and its formation occurs over the crystalline MoO₃ phase. The Selectivities to CO and methyl formate are greatly enhanced because of the presence of support, and are relatively independent of the catalyst loading and phase distribution. The dependence and independence of the Selectivities of different byproducts on the loading make the silica supported catalysts with high catalyst loadings less selective for the partial oxidation of methanol to formaldehyde.