摘要
提出一种基于HOG特征的学生课堂状态图像检测方法,从课堂监控图像中提取反映课堂教学效果与学生课堂状态的相关信息,为后续行为状态识别提供基础数据。首先,截取包含学生完整课堂状态的图像块,提取其HOG特征;然后,运用Adaboost分类算法训练级联分类器;最后使用图像金字塔技术滑动检测,检出学生听课状态窗口。
This paper proposes a detection method of students’ classroom states based on HOG features,by extracting the relevant information reflecting classroom teaching effect and students’ classroom state from classroom monitoring images to provide data foundation for subsequent behavior status recognition. Firstly,intercept the image blocks containing a student’s complete classroom status,extract its HOG features;then use Adaboost classification algorithm to train the cascaded classifier;finally the image pyramid technology is used to detect the status window of the student’s lecture.
作者
陈杰
谢日敏
CHEN Jie;XIE Rimin(Information Technology Center of Fujian Business College, Fuzhou 350012, China;Department of Information Engineering, Fujian Business College, Fuzhou 350012, China)
出处
《重庆科技学院学报（自然科学版）》
CAS
2019年第4期89-92,共4页
Journal of Chongqing University of Science and Technology：Natural Sciences Edition
基金
福建省中青年教师教育科研项目“基于校园视频监控的物体检测识别关键技术研究”(JT180636)
