摘要
针对传统的SIFT人脸识别算法存在特征维数高、匹配计算量和难度大的问题,采用OCR技术中的非均匀网格的方式对人脸图像进行区域划分,引入旋转无关的等价模式的LTP特征对SIFT关键点进行描述并对比实验.研究结果表明:改进后的SIFT人脸识别算法降低了特征的维数,增加了对旋转、光照变化、噪声干扰等影响因素的鲁棒性.研究结论初步突破了传统SIFT人脸识别算法,有助于从OCR技术中寻求提高识别率,降低匹配计算的复杂程度的方法.
In view of the traditional SIFT face recognition algorithm,which has high feature dimension,matching calculation amount and difficulty,the non-uniform grid method in OCR technology is used to divide the face image,and the rotation-independent equivalent mode is introduced.The key points of SIFT with the LTP features is described and comparative experiments are carried out.Experimental results show that the improved SIFT face recognition algorithm not only reduces the dimensionality of features,but also increases robustness to factors such as rotation,illumination changes,and noise interference.The research conclusions have made a preliminary breakthrough in the traditional SIFT face recognition algorithm,which helps to find ways to improve the recognition rate and reduce the complexity of matching calculations from OCR technology.
作者
霍春宝
杨闯
佟智波
杨红喆
王丹丹
HUO Chunbao;YANG Chuang;TONG Zhibo;YANG Hongzhe;WANG Dandan(School of Electrical Engineering,Liaoning University of Technology,Jinzhou 121001,China;Mining Business Division,Jinzhou Petrochemical Company,Jinzhou 121001,China;Power Dispatching Control Center,State Grid Jinzhou Power Supply Company,Jinzhou 121001,China;School of Computer Science and Information Engineering,Shanghai Institute of Technology,Shanghai 200235,China)
出处
《辽宁工程技术大学学报（自然科学版）》
CAS
北大核心
2021年第4期378-382,共5页
Journal of Liaoning Technical University (Natural Science)
基金
辽宁省高校重大科技平台项目(JP2017011)
作者简介
霍春宝(1972-),男,辽宁锦州人,博士,教授,主要从事智能控制与信息处理方面的研究.
