视频人脸图像质量评估及其在课堂点名系统中的应用

FACE IMAGE QUALITY EVALUATION IN VIDEO STREAM AND ITS APPLICATION IN CLASSROOM ATTENDANCE SYSTEM

班级课堂考勤是课堂管理的有效手段之一,可以有效地监督学生按时上课,保证课堂的教学质量。近年来,深度学习在静态图像人脸识别方面已经取得较大进展。但在课堂环境下的视频流中,课堂环境人脸位置不一、人体不停运动和姿态偏移较大等现象会导致关键帧中检测到的部分人脸区域存在人脸尺寸较小、运动模糊与像素低等问题。在传统人脸识别系统中引入人脸图像质量评估方法。通过该评估方法自动筛选出关键帧中人脸特征明显的图像,以保证人脸识别系统在课堂环境中的有效性和鲁棒性。实验结果表明,该人脸质量评估方法可以准确地过滤人脸特征不明显的关键帧,有效地提高了人脸识别系统的准确率,大大提高了课堂点名的效率。

Class attendance is one of the effective means of classroom management, which can effectively supervise students to attend class on time and ensure the quality of classroom teaching. In recent years, deep learning has made great progress in static image face recognition. However, in the classroom video stream, the phenomena that face position is different, human body keeps moving and posture deviation is larger in classroom environment will lead to several problems that some facial regions detected in key frames have smaller face size, motion blur and lower pixels. We introduced a face image quality evaluation method in the traditional face recognition system. Through this evaluation method, the images with obvious facial features in key frames were automatically screened out to ensure the effectiveness and robustness of face recognition system in classroom environment. The experimental results show that the proposed method can accurately filter the key frames with inconspicuous facial features, effectively improve the accuracy of face recognition system, and greatly improve the efficiency of classroom attendance.