基于OpenStack构建手写识别数据即服务云平台

发布时间：2018-08-02 14:43

【摘要】：近年来，触屏的普及大大推进手写交互的应用，手写识别技术日显其价值。同时，行业信息化建设的进程开始面临大数据的挑战，包括了数据的运维、数据资源的重分配、数据向业务价值的转化等，，从而产生了行业云理念。商业和学术界的许多行业都面临着这样的调整，本文所研究的手写识别领域也同样面临行业大数据的机遇和挑战。因此，本文针对手写识别研究领域，提出一种行业云的解决方案。本文提出了基于开源云计算平台OpenStack IaaS层构建的手写识别数据即服务云平台（Handwriting Recognition Data Platform as a Service，简称：HRDPaaS）。其目的和意义在于为解决手写识别当前遇到的问题提供合理的方案：1）低配置端（如嵌入式、Web端）的手写识别难以实现；2）个性化的手写识别运算复杂，需要大存储、高配置；3）手写识别研究所需的样本集成本很高、数量很大；4）该领域数据资源不均衡、数据服务缺失。本文所提出的平台针对手写识别的技术领域，以手写数据为核心，开放手写笔迹识别服务、手写数据统计服务、手写样本调度服务、数据可视化服务等。同时对接平台基础资源监控系统以及数据运维系统，利用基于OpenStack的弹性伸缩、负载均衡等解决方案，实现平台计算资源的部署和调度。本研究课题实现的系统和服务包括：1）在线手写识别，拓宽了手写交互的应用场合；2）大量用户的海量数据存储，为个性化手写识别的研究提供数据基础；3）提供大量的样本调度，降低手写识别研究的成本；4）整合了海量具有自然语义的样本集，为手写样本数据向语义的转化提供基础。利用该平台，可以通过数据，整合、优化、重组手写交互应用和研发的生态链。
[Abstract]:In recent years, the popularity of touch screen has greatly promoted the application of handwritten interaction, and handwriting recognition technology is becoming more and more valuable. At the same time, the process of the construction of industry information began to face the challenge of big data, including the operation and maintenance of data, the redistribution of data resources, the transformation of data to business value, and so on, which resulted in the concept of industry cloud. Many industries in business and academia are facing such adjustment. The handwritten recognition field studied in this paper also faces the opportunities and challenges of industry big data. Therefore, this paper proposes a solution of industry cloud for handwritten recognition. This paper proposes a handwritten recognition data called Service Cloud platform (Handwriting Recognition Data Platform as a Service,) based on the open source cloud computing platform (OpenStack IaaS layer). Its purpose and significance is to provide a reasonable scheme for solving the current problems encountered in handwriting recognition: 1) it is difficult to realize handwriting recognition in low configuration terminal (such as embedded web side). 2) the operation of personalized handwriting recognition is complex, which needs large storage and high configuration. 3) the cost of the sample set is very high and the quantity is very large. 4) the data resource in this field is unbalanced and the data service is missing. Aiming at the technical field of handwritten recognition, the platform proposed in this paper is based on handwritten data, and open handwritten handwriting recognition service, handwritten data statistic service, handwritten sample scheduling service, data visualization service and so on. At the same time, the basic resource monitoring system and the data operation and maintenance system of the platform are docked to realize the deployment and scheduling of the platform computing resources by using the solutions of elastic expansion and load balancing based on OpenStack. The system and service implemented in this research include: 1) online handwriting recognition, which broadens the application of handwritten interaction and broadens the mass data storage of a large number of users, and provides a data basis for the study of personalized handwriting recognition. 3) providing a large number of samples scheduling and reducing the cost of handwritten recognition. 4) integrating a large number of samples with natural semantics to provide the basis for the transformation of handwritten sample data to semantics. With this platform, the ecological chain of handwritten interactive applications and R & D can be reorganized through data, integration, optimization and reorganization.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.41;TP393.09

【参考文献】