现代电子技术2018，Vol.41Issue(6)：50-52,56,4.DOI:10.16652/j.issn.1004-373x.2018.06.012
基于多核学习SVM的图像分类识别算法
An image classification and recognition algorithm based on multi-kernel learning SVM
摘要
Abstract
In allusion to the problem that the single kernel support vector machine(SVM)has low detection precision in image classification and recognition,an image classification and recognition algorithm based on multi-kernel learning SVM is proposed and applied to the pedestrian detection problem. The pedestrians′ integral channel features including the gradient histo-gram,the color channel and the gradient feature are extracted. The hybrid kernel SVM classifier is constructed by using the his-togram cross kernel,the polynomial kernel,and the radial basis kernel. The methods of cross validation and grid search are used to determine the fusion coefficients of various kernels. The results from the TUD dataset test show that the method has good robustness and high detection precision.关键词
支持向量机/多核学习/行人检测/图像识别/直方图交叉核/交叉验证Key words
SVM/multi-kernel learning/pedestrian detection/image recognition/histogram cross kernel/cross validation分类
信息技术与安全科学引用本文复制引用
李红丽,许春香,马耀锋..基于多核学习SVM的图像分类识别算法[J].现代电子技术,2018,41(6):50-52,56,4.基金项目
河南省科技厅科技攻关项目(172102210604 （）
162102210131) （）
河南省高等学校重点研究项目(16B510008) （16B510008）
中州大学科技创新团队建设计划项目(CXTD2017K4) （CXTD2017K4）
郑州工程技术学院教育教学改革研究项目(16) Project Supported by Key Scientific and Technological Project of Henan Provincial Science and Technology Department(172102210604,162102210131), Key Research Project of Colleges and Universities in Henan Province(16B510008),Construction Plan of Science and Technology Innovation Team in Zhongzhou University(CXTD2017K4),Education and Teaching Reform Research Project of Zhengzhou Institute of Technology(16) （16）
