We construct a call network digraph G with attribution using mobile phone call records within 21 days collected by three operators and analyze the social call behavior features through analyzing the degree distribution of nodes of the network digraph under the help of Graphx based on the Spark Platform.We get the following social conclusions from the communication relationship between phone users:on average,users make about5~7 mobile telephone calls and connect about2~3 persons one day;on weekends,people make less calls but spend longer time on each call,revealing that working calls account for a large part of calls on weekdays;among these telephone calls on weekdays,most are less than one minute,and on average,mobile users that call more people also tend to be called by more individuals.
Changyuan WangYang LiuZongfei LuShanqing GuoLizhen Cui
目的纹理是描述和区分不同物体的重要特征之一,纹理特征提取一直是模式识别、机器视觉领域的研究热点。局部方向模式(LDP)是一种分辨性好、对随机噪声和非均匀光照鲁棒的纹理特征。而LDP特征由于计算8方向的边缘响应并排序,提取速度较慢。为此对LDP编码方案进行改进。方法设计了两种改进方案:第1种方案直接对8方向的边缘响应符号进行编码,避开排序,称为FLDP(fast local directional pattern)特征;第2种方案,尝试使用较少的方向模板来降低特征提取的时间、空间消耗,设计了MLDP算子(mini local directional pattern)。结果在Brodatz数据集的24类均匀纹理图像以及111类全部纹理图像上将本文提出的FLDP特征、MLDP特征与传统的LDP进行了对比实验。实验结果表明,在保证了分类准确率的前提下,FLDP算子的运算速度是3th-LDP的20倍左右,MLDP算子的运算速度是3th-LDP的35倍左右。结论论文设计了2种方案改进了LDP特征,分别为FLDP算子和MLDP算子。实验结果表明,这两种改进方案,在保证分类准确率的同时,大幅度提高了特征提取运算速度。
