摘 要:随着深度学习技术的不断发展,端到端的模式识别方式越来越普及。只要算力足够,总可以搭建出适合某一场景的分类检测与预测判断深度网络。但在周期短、算力不足、语义理解要求高的场景下,基于特征的模式识别仍旧有着巨大的需求。本文通过对经典特征HOG 的详细解读,结合SVM 实现了对任意物体的智能识别。相较于深度学习算法,有着前期训练成本低、识别速度快、样本量需求小等特点。
关键词:模式识别;SVM;HOG;目标检测;人工智能
中图分类号:TP391.41 文献标识码:A 文章编号:2096-4706(2019)24-0067-04
A Common Object Recognition Method Based on HOG+SVM
LI Ming,ZHENG Susheng,YAO Leiyue
(Jiangxi University of Technology,Nanchang 330098,China)
Abstract:With the development of deep learning technology,the end to end method has become the most popular pattern recognition method nowadays. In most situations,an appropriate deep learning network always can be created with the help of enough computing power. However,in the scene of short project period,insufficient computing power,high competence of algorithm etc.,feature based pattern recognition method still has its application requirements. In this paper,the classic feature,HOG was fully introduced,which together with SVM(support vector machine) realized a common algorithm for any certain object detection. Compared with deep learning algorithm,it has the characteristics of low training cost,fast recognition speed and small sample size demand.
Keywords:pattern recognition;SVM;HOG;object detection;artificial intelligence
基金项目:本文系2018 年度江西省社会科学规划项目:基于教学过程大数据的民办高校教学资源配置优化研究——以江西科技学院为例(项目编号:18JY40);2017 年江西省科技厅工业领域科技攻关项目:基于自然语音交互模式的行车安全辅助系统(项目编号:20171BBE50060);南昌市科技局指导性科技计划项目:汽车分时租赁远程控制与准入系统(项目编号:GJJ161143)资助。
参考文献:
[1] DALAL N,TRIGGS B. Histograms of Oriented Gradientsfor Human Detection [C]//2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’05). IEEE,2005.
[2] 周志华. 机器学习 [M]. 北京:清华大学出版社,2016:121-139+298-300.
[3] LI Z,ZHOU F. FSSD:Feature Fusion Single Shot MultiboxDetector [J/OL].arXiv,2017(14):1-10(2017-12-04),https://arxiv.org/pdf/1712.00960v1.pdf.
[4] 阳冰成. 视频序列中行人目标检测与跟踪 [D]. 北京:北京印刷学院,2019.
[5] 韩涛,杨洋. 基于Hough 变换的图像目标检测与识别 [J].计算机与数字工程,2019,47(2):412-416.
作者简介:
李铭(1985.07-),男,汉族,江西抚州人,助教,本科,研究方向:计算机网络;
郑苏生(1978.01-),男,汉族,江西抚州人,讲师,研究生,研究方向:工商管理;
姚磊岳(1982.07-),男,汉族,江西南昌人,教授,研究生,研究方向:大数据挖掘与人工智能。