基于Hadoop的舌部图像预处理时间对比研究

摘要

In order to compare the tongue image preprocessing time of the single machine system and the Hadoop based system, 1482 cases of standardized tongue image are collected from the medical examination center of Tianjin Nankai hospital. These images are pre-processed by a single PC, a two-worker and a four-worker Hadoop system respectively. Three kinds of system run the preprocessing for 5 times each, average the processing time of each system then contrast. Experimental results show that the two-worker system is reduced to 52.1% compared with the single machine system, and the four-worker system is reduced to 28.1%. Compared to parallel computer, using the existing computer and network resources, the tongue image preprocessing system based on multi-worker Hadoop can effectively shorten the preprocessing time while the costs are not significantly increased.%为了对比单机系统和基于Hadoop系统的舌部图像预处理所用时间，从天津南开医院体检中心采集了1482例标准化舌部图像。使用双Worker、四Worker的Hadoop系统与单机系统对这些图像进行预处理。三种系统下运行预处理各5次，对处理时间取平均值并对比。实验结果，表明双Worker系统所用时间与单机系统相比缩短到了52.1%，四Worker系统缩短到了28.1%。相对于并行计算机，基于Hadoop多Worker的舌部图像预处理系统利用现有计算机和网络资源，在几乎不增加成本的情况下有效地缩短了预处理时间。