Aroma Impact Compounds in Three Citrus Oils: Cross-matching Test and Correspondence Analysis Approach

ABSTRACT Twenty‐three odor chemicals from 3 Citrus essential oils (lemon, Valencia orange, and Citrus sudachi) were selected as the potent character‐impact compounds on the basis of their limited odor unit values, and then every chemical was cross‐matched by sensory test to the 3 oils to attribute each aroma character to 1 of the 3 Citrus oils. The matching‐frequency data (ratio data) obtained was subjected to correspondence analysis and graphed on a diagram. Consequently, it was found that the aroma character of lemon oil was mainly represented by citral, with a high matching frequency of 0.89 (59 counts out of 66 trials, x²_o= 93.36). The orange character consisted mostly of linalool and nonanal. α, β‐Pinene, α‐sinensal and myrcene were related to the aroma of C. sudachi oil. The application of the cross‐matching test and correspondence analysis in the characterization of food aromas has never been reported in the literature to date, and the validity of these methods was successfully demonstrated by our study.     

前背景骨架区域随机点对策略驱动下的车标识别方法

目的 现有的车标识别算法均为各种经典的图像特征算子结合不同的分类器组合而成,均未分析车标图像的结构特点。综合考虑车标图像的灰度特征和结构特征,提出了一种前背景骨架区域随机点对策略驱动下的车标识别方法。方法 本文算法将标准车标图像分为前景区域和背景区域,分别提取前、背景的骨架区域,在其中进行随机取点,形成点对,通过进行点对的有效性判断,提取能表示车标的点对特征。点对特征表示两点周围局部区域的相似关系,反映了实际车标成像过程中车标图案部分与背景部分的灰度明暗关系。结果 在卡口系统截取的19 044张车标图像上进行实验,结果表明,与其他仅基于灰度特征的识别方法相比,本文提出的点对特征识别方法具有更好的识别效果,识别率达到了95.7%。在弱光照条件下,本文算法的识别算法效果同样优于其他仅基于灰度特征的识别方法,识别率达到了87.2%。结论 本文提出的前背景骨架区域随机点对策略驱动下的车标识别方法,结合了车标图像的灰度特征和结构特征,在进行车标的描述上具有独特性和排他性,有效地提高了车标的识别率,尤其是在弱光照条件下,本文方法具有更强的鲁棒性。     

用MapReduce实现天文星表交叉认证*

天文星表的交叉认证是天文研究中非常重要的基础工作。新巡天项目和更强大望远镜的投入使用,使天文数据爆炸增长,数据量的增加使得两个星表之间的交叉认证变得非常耗时。描述了如何利用MapReduce实现并行天文星表交叉认证,综合考虑了算法与体系结构的匹配问题,并给出了在大数据天文星表交叉认证工作的性能评估,通过与广泛使用的PostgreSQL数据库的比较,证明了基于MapReduce交叉认证方法的有效性。