Research On The Evaluation Method Of Classroom Concentration Based On Multi-information Fusion

Learning concentration is one of the key factors affecting the quality of teaching.Concentration evaluation is in line with the needs of education and teaching reform,and is conducive to the improvement of teaching quality.Information technologies such as artificial intelligence and computer vision have accelerated the intelligent development of focus assessment.When assessing the concentration of students in a classroom scene,it is difficult to extract detailed information due to the low resolution,occlusion,and uneven lighting of images captured by classroom monitoring equipment,which brings certain challenges to the concentration assessment work..In order to realize the learning concentration evaluation of dense student groups in classroom scenarios,combined with the teaching space-time structure in educational sociology,position information,head posture and movement frequency are used as evaluation factors,and a classroom concentration assessment method based on multi-information fusion is designed.The main work is as follows:1.A method of location information detection is designed.The analysis of the space-time structure of classroom teaching needs to be accurate to the spatial position of each student in the classroom,and the description of the spatial position often depends on the seat position.Affected by the shooting angle of the camera,the positional relationship between the student and the seat is different in the image coordinate system and the space coordinate system.In order to accurately detect and describe the student’s position information in the image,the HSV color model is used to detect the seats in idle classrooms.This method effectively solves the color difference of the seats caused by the influence of light,which is better than the RGB-based color detection method.Then,the Face Boxes network is used for face detection to obtain the location information of students,which solves the problem of detecting small faces in the back row of the classroom.Finally,by dividing the student’s front face and side face area,and combining the Manhattan distance and the Euclidean distance to match the student’s face and the seat,the accurate description of the student’s position in the classroom is completed.2.The head pose discrimination and motion frequency calculation methods are designed.Head posture and movement frequency determine students’ interest and enthusiasm in teaching activities,which reflects students’ concentration level.Affected by the image resolution,it is difficult to extract facial details to obtain Euler angles to estimate the head pose.Therefore,the head-up and head-down state discrimination is realized by combining position information,face detection confidence rate and aspect ratio.Since there is occlusion in the rear row,motion detection cannot be performed,so the frame difference method is used to detect the contour of the moving target,and the faces contained in the contour are counted.After matching the face and the seat,the motion frequency statistics are completed.Referring to the teaching space-time structure,the rear row refers to the front row for head posture discrimination and motion frequency calculation.This method makes full use of limited information to maximize the extraction of effective features.3.A comprehensive discrimination method of concentration based on multi-information fusion is designed.The evaluation factors obtained above are divided into 7 categories: front row,rear row,head up,head down,high frequency of exercise,moderate frequency and low frequency.The self-adjusting AHP is used to determine various weights,which improves the scientificity of weight allocation.Combined with the weighting method to achieve the quantification of multi-information,the learning concentration evaluation is completed by dividing very focused,general focused and unfocused.The concentration discrimination method can objectively evaluate students’ learning concentration level,which has certain practicability,and provides a method reference for concentration evaluation in other scenarios.