摘要
在移动终端进行商标识别时,使用图像处理技术的传统方法已不再适用.针对这一问题,提出基于计算迁移的智能终端商标识别方法.首先通过分析商标识别应用的组织架构来划分商标识别应用的任务节点;然后构建基于任务节点的应用成本图,设计计算迁移的约束条件;最后制定计算迁移的目标函数,以实现应用能耗与应用时长的均衡.实验结果表明,该方法可以提升移动终端商标识别应用的性能,降低本地能耗,增强用户体验.
Trademark is the mark that distinguishes enterprise brand or service, and plays an important role in spreading enterprise culture. Trademark recognition application on intelligent terminal is a computationally intensive application. The application performance is limited by the resource bottlenecks of intelligent terminal. Address to the problem, this paper presents a computation offloading method for trademark recognition application on intelligent terminal. In this method, the entire application is firstly divided into many task nodes by categorizing their usage. Then the energy consumption and execution time of each single task node are monitored and calculated, and an application cost graph can be built up. Finally, the objective functions based on minimum energy consumption, minimum executed time, and the balance between energy consumption and executed time can be derived. In the paper, offloading experiments based on these objective functions are carried out, and the experimental results show that the proposed method can improve the performance of trademark recognition application on intelligent terminal, reduce the energy consumption, and enhance user experience.
作者
张金
冯帆
孙承君
宫晓利
ZHANG Jin;FENG Fan;SUN Cheng-jun;GONG Xiao-li(College of Computer and Control Engineering,Nankai University,Tianjin 300350,China;State Key Laboratory of Computer Architecture,Institute of Computing Technology,Chinese Academy of Sciences,Beigjing 100190,China)
出处
《应用科学学报》
CAS
CSCD
北大核心
2018年第4期667-678,共12页
Journal of Applied Sciences
基金
国家自然科学基金(No.61702286)
体系结构国家重点实验室开放课题基金(No.CARCH201604)
天津市优秀科技特派员基金(No.17JCTPJC49500)资助
关键词
计算迁移
商标识别
云计算
目标函数
computation otItoad
trademark recognition
cloud computation
object function
