基于MapReduce的人脸识别的研究
发布时间:2018-10-09 18:26
【摘要】:随着大数据时代和智能化的到来,如何高效的从人脸图像数据中挖掘出有价值的信息,已经成为人脸识别领域研究的热点。随着云计算、分布式批处理的飞速发展,给人脸识别带来了新的思路。传统人脸识别技术只针对小范围内、静止状态下、单个人脸识别的研究。通常在数据量增大的时候实时性非常低,因此无法使用在数据量较大的场所。本文对人脸识别在适用范围和实时性进行探索,提出了将现有的人脸识别算法和大数据处理架构中的批处理计算框架MapReduce相结合的新思路。本文的主要贡献如下:1)针对传统人脸识别适用范围小的问题,提出了使用大数据存储技术Hadoop中的HDFS和HBase来存储数据。将所有的内置图片和待识别的人脸图片存放在HBase中,将唯一代表内置图片信息的文本文件存放在HDFS中,进而使得人脸识别可以应用于范围更大的场所。2)针对传统人脸识别实时性低的问题,提出了将人脸识别算法和Hadoop中的批处理MapReduce相结合的思路。首先对人脸识别PCA算法,使用Map计算欧式距离,得到处理的中间的结果。然后用Reduce处理此中间结果,最后将得到的最小欧式距离所对应的内置图片信息作为最终结果并存储。为测试本文改进的人脸识别系统的性能,用多组不同数量的内置人脸图片数据来评估对人脸识别模型改进的实时性和适用范围,得到了良好的测试结果。
[Abstract]:With the era of big data and the arrival of intelligence, how to efficiently extract valuable information from face image data has become a hot spot in the field of face recognition. With the rapid development of cloud computing, distributed batch processing has brought new ideas to face recognition. The traditional face recognition technology only focuses on single face recognition in a small range and static state. Generally, the real-time performance is very low when the amount of data increases, so it can not be used in places with large amount of data. This paper explores the applicability and real-time of face recognition, and proposes a new idea of combining the existing face recognition algorithm with the batch computing framework (MapReduce) in big data's processing architecture. The main contributions of this paper are as follows: (1) aiming at the problem of small application range of traditional face recognition, this paper proposes to use HDFS and HBase in big data storage technology to store data. All the built-in images and face images to be recognized are stored in the HBase, and the unique text files representing the built-in picture information are stored in the HDFS. So that face recognition can be applied to a larger range of places. 2) aiming at the problem of low real-time performance of traditional face recognition, the idea of combining face recognition algorithm with batch processing MapReduce in Hadoop is proposed. Firstly, the Euclidean distance is calculated by using Map to calculate the Euclidean distance for face recognition PCA algorithm, and the intermediate results are obtained. Then the intermediate result is processed by Reduce, and the built-in picture information corresponding to the minimum Euclidean distance is finally stored as the final result. In order to test the performance of the improved face recognition system in this paper, the real-time and applicable range of the improved face recognition model is evaluated by using a number of different sets of built-in face image data, and a good test result is obtained.
【学位授予单位】:西安科技大学
【学位级别】:硕士
【学位授予年份】:2017
【分类号】:TP391.41
本文编号:2260329
[Abstract]:With the era of big data and the arrival of intelligence, how to efficiently extract valuable information from face image data has become a hot spot in the field of face recognition. With the rapid development of cloud computing, distributed batch processing has brought new ideas to face recognition. The traditional face recognition technology only focuses on single face recognition in a small range and static state. Generally, the real-time performance is very low when the amount of data increases, so it can not be used in places with large amount of data. This paper explores the applicability and real-time of face recognition, and proposes a new idea of combining the existing face recognition algorithm with the batch computing framework (MapReduce) in big data's processing architecture. The main contributions of this paper are as follows: (1) aiming at the problem of small application range of traditional face recognition, this paper proposes to use HDFS and HBase in big data storage technology to store data. All the built-in images and face images to be recognized are stored in the HBase, and the unique text files representing the built-in picture information are stored in the HDFS. So that face recognition can be applied to a larger range of places. 2) aiming at the problem of low real-time performance of traditional face recognition, the idea of combining face recognition algorithm with batch processing MapReduce in Hadoop is proposed. Firstly, the Euclidean distance is calculated by using Map to calculate the Euclidean distance for face recognition PCA algorithm, and the intermediate results are obtained. Then the intermediate result is processed by Reduce, and the built-in picture information corresponding to the minimum Euclidean distance is finally stored as the final result. In order to test the performance of the improved face recognition system in this paper, the real-time and applicable range of the improved face recognition model is evaluated by using a number of different sets of built-in face image data, and a good test result is obtained.
【学位授予单位】:西安科技大学
【学位级别】:硕士
【学位授予年份】:2017
【分类号】:TP391.41
【参考文献】
相关期刊论文 前9条
1 赵士伟;张如彩;王月明;张晖;;生物特征识别技术综述[J];中国安防;2015年07期
2 阮梦黎;;大数据挑战下的NoSQL系统研究[J];聊城大学学报(自然科学版);2015年01期
3 李青云;余文;;关系型数据库到H Base的转换设计[J];信息网络安全;2015年01期
4 李伟;;Hadoop平台下的分形图像压缩编码[J];测控技术;2014年04期
5 卢世军;;生物特征识别技术发展与应用综述[J];计算机安全;2013年01期
6 张会森;王映辉;;人脸识别技术[J];计算机工程与设计;2006年11期
7 张晓华,山世光,曹波,高文,周德龙,赵德斌;CAS-PEAL大规模中国人脸图像数据库及其基本评测介绍[J];计算机辅助设计与图形学学报;2005年01期
8 张新宇,张三元,叶修梓;基于图像化几何的三维模型彩绘[J];软件学报;2004年03期
9 张翠平,苏光大;人脸识别技术综述[J];中国图象图形学报;2000年11期
相关硕士学位论文 前1条
1 张臻;基于Hadoop的并行优化方法及其在人脸识别中应用研究[D];电子科技大学;2015年
,本文编号:2260329
本文链接:https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2260329.html