Research On Classroom Behaviors Recognition Method Based On Digital Image Processing

Teaching in classroom is a basic form of teaching,and it is the main channel for students to obtain information,improve skills and form ideas.The teaching and learning behaviors of teachers and students have a great influence on the educated.These behavioral data can be used to objectively evaluate students,predict students' future academic grades,plan students' occupations,evaluate teachers' teaching level,push teachers' introspection and improve quality of teaching.Therefore,to recognize behaviors of teaching and learning has a very wide application and is deserved to be studied in depth.At present,the research on behaviors recognition in the complex classroom environment has just been carried out.This thesis studies methods to recognize standing and raising hands in classroom based on digital image processing.The main contents of this thesis are as follows:(1)The realization and application of the moving target detection algorithms in classroom.The typical moving target detection algorithms,including Motion History Image,Gaussian mixture model,Vibe,Optical flow method,are applied to the classroom,and the experimental results are analyzed.Due to the lack of foreground extraction of these algorithms,the difficulties still exist for classroom behavior detection.A idea is determined that first detects the subject(face,hand)and then judges the subject's movement tendency.(2)Standing recognition based on the region of interest.In order to avoid the influence of the background noise of the characters,combined with the characteristics of the main motion occurring in the upper part of the image,the dichotomy method is used to determine the clipping position,the image is cropped,the region of interest is obtained,and different thresholds are selected according to whether there is any interference with hands.The segmentation algorithms are applied to recognize standing under different segmentation backgrounds.(3)Standing recognition based on face tracking.In order to locate and close up the behavioral target,the Viola-Jones face detection algorithm is used to detect the face and frame it.A block-based inheritance algorithm based on position and area is proposed.The algorithm uses closest principle of the old and the new frames based on position and area which are basically unchanged between close frames.Through the bidirectional selection of the old and new frames,the corresponding relationship between the frames is determined,and the face tracking is completed.By analyzing the coordinate change of face,the movement trend of face is judged,and the recognition of stand-up behavior is realized.(4)Raising hands recognition based on skin color detection and hand tracking.Aiming at the color of the hand,most of the non-skin area is removed by the skin color detection algorithm.In order to remove the face interference,a scan line seed removal algorithm is proposed by using the holes formed by the facial features in the face,and all the connected domains containing the holes are removed.That is,the face is removed.The contour of the hand is found by the boundary tracking algorithm.The position of the hand is determined by the pole information.The recognition of raising hands is also realized by the block diagram inheritance algorithm similarly.