Attention Detection Technology Based On Visual Features Fusion And Its Research On Online Teaching

Online teaching is popular with the ubiquitous internet today.During the COVID-19 pandemic,people’s demand for non-contact online teaching has become more urgent.Online teaching without time and space constraints makes people’s learning more free and convenient.However,compared with the traditional classroom,the interaction between teachers and students in the online classroom environment is poor,and there are many obstacles in communication.It is difficult for teachers to effectively monitor students’ learning status,and students may also get lower learning efficiency due to lack of supervision.With the characteristics of online teaching environment are considered,this paper proposes an attention detection method based on visual feature fusion to alleviate the above issues,which includes attention detection related technologies such as no-person detection,head deflection detection,eye closure detection and eye gaze deviation detection.Finally the author designs and develops an attention detection system.The main contents include:(1)An iris center localization algorithm based on snakuscule is proposed.In order to improve the iris center localization accuracy of low-resolution face images acquired under natural light,an improved snakuscule energy model is proposed,and an adaptive iteration strategy is designed,combined with single-pixel iteration,skip iteration and the proportion of iris inner circle.The iteration modes are increased to10 to improve the real-time performance and accuracy.Then,iris quality inspection is introduced to decrease the detection error and improve the localization accuracy.Compared with other iris center localization algorithms,the superiority and accuracy of the proposed algorithm are verified.(2)In order to improve the accuracy of gaze estimation under different poses,a gaze estimation algorithm is proposed.Two types of eye ROI segmentation methods are modeled and analyzed respectively,and a targeted scheme is given to deal with the gaze estimation under the three postures of face forward,face tilt and face deflection.The iris center compensation algorithm and main eye strategy are introduced to reduce the impact of head deflection on the focus of sight.Finally,experiments are carried out based on this algorithm,whose results show the effectiveness and accuracy of this scheme.(3)Integrating multiple attention related features,an attention detection system incorporating multiple features is designed and developed.Combined with a variety of visual features,this author designs and develops four attention detection modules,including no-person detection,head deflection detection,eye closure detection and eye gaze deviation detection.The system is composed by student-side subsystem and teacher-side subsystem.The student-side subsystem can transmit the attention detection results to the teacher-side subsystem,while teachers can pay real-time attention to the learning status of all online students through the teacher-side subsystem.Experiment is carried out to compare the accuracy of attention detection based on the fusion of single visual features and that of multiple visual features.The results show that the detection system implemented in this paper can accurately complete the task of attention detection,so as to improve learning efficiency and teaching quality.