一种基于视觉的图书馆占座检测方法  

A Seat Occupancy in the Library Detection Method Based on Vision

作　　者：王建功[1] 林国余 WANG Jiangong;LIN Guoyu(Library, Minjiang University, Fuzhou, Fujian 350108, China;Nanjing Zhizhen Electronic Technology Co., Ltd, Nanjing, Jiangsu 210096, China)

机构地区：[1]闽江学院图书馆,福建福州350108 [2]南京智真电子科技股份有限公司,江苏南京210096

出　　处：《闽江学院学报》2022年第2期52-60,共9页Journal of Minjiang University

摘　　要：为解决图书馆阅览室经常存在的严重占座现象,提出一种基于图像处理的图书馆占座检测方法。在每个座位上方安装一个光轴垂直朝下的摄像机,采用Hough方法实时检测桌子区域;通过背景建模方法检测前景目标,并提取目标的HOG特征,通过事先训练好SVM分类器判定目标为人体还是位于桌面上的其它目标。在此基础上,根据桌子座位先验知识和人体数量,判定是否存在占座现象。通过图书馆实际场景的测试,验证了该方法能够有效检测桌子是否存在占座行为。In order to solve the frequent occurrence of serious seat occupancy in the reading room of the library,a library seat occupancy detection method based on image processing is proposed.In the method,a camera is mounted over every reading table and its axis is perpendicular to the ground.Firstly,the Hough method is used to detect the table area.Secondly,the background modeling approach is adopted to detect the foreground objects and then the HOG features of the objects are acquired.Then,the SVM classifier trained previously is exploited to detect the human heads and other things on the table.Based on that,according to the human counts,prior knowledge of the seat number and the location of the things on the table,seat occupancy can be inferred.The experiments in the library prove that the proposed method can detect the seat occupancy effectively.

关 键 词：占座检测 视觉检测 梯度直方图 支持向量机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]