The Design And Implementation Of The Real-Time Super-Resolution Scheme Of The Blackboard Image In The Online Classroom Scene

With the development of Internet technology,online education has gradually become one of the ways for people to learn,and the online classroom live teaching mode has attracted attention due to its real-time interactive function.This model not only provides opportunities for teachers and students to communicate remotely,but also helps more students receive highquality educational resources in a timely manner.However,the current online classroom live teaching mode still has the defect that the real-time performance and the video picture quality cannot be satisfied at the same time.According to the application scenario of online classroom live teaching mode,a real-time detection and reconstruction scheme is designed based on blackboard screen.Firstly,considering that the core of the video image in this scenario is the blackboard writing area,this scheme uses the object detection algorithm Yolov5 n to extract this area to reduce the area for subsequent reconstruction.Secondly,the video is compressed to ensure real-time transmission,but the quality of the video is compromised.Therefore,a real-time video communication superresolution network based on deep learning,RVCSRN,is proposed.In addition,combined with the feature that high-resolution images can be directly obtained in online classroom scenarios,this scheme adopts the reconstruction network DI-RVCSRN in dual-input mode in the reconstruction stage.The main work of this paper is as follows:(1)Requirements analysis and solution design.Firstly,a requirements analysis is carried out for the online classroom live teaching task,specifying the two needs of meeting real-time and improving video quality.Then,two methods of target detection and super-resolution are used to design targeted solutions.Finally,the corresponding blackboard datasets are homemade according to the characteristics of the target scene.(2)Determine the blackboard writing area detection algorithm.For the detection task of the blackboard regions,the mainstream two-stage detection algorithm Faster R-CNN and the single-stage detection algorithm Yolov5 n were experimentally compared based on the self-built blackboard frame dataset,and after comprehensively considering the advantages and disadvantages of these algorithms,Yolov5 n was finally selected as the blackboard region detection algorithm.(3)Design the reconstruction algorithm for blackboard writing.In order to ensure the realtime performance,the down-sampling method is used at the send end to compress the amount of transmitted data,and the super-resolution algorithm is used at the receiving end to improve the quality of the damaged picture.Aiming at the super-resolution reconstruction of blackboard regions,the advantages and disadvantages of the existing methods are pointed out,and the realtime video communication super-resolution network RVCSRN is proposed by using the improved enhanced information multi-distillation block EIMDB and pixel-level information distillation block PIDB.In addition,since the online classroom scene can directly obtain highresolution pictures,a dual-input mode reconstruction method is adopted in the reconstruction stage,and the DI-RVCSRN network is proposed.Then,comparative experiments are designed on public datasets to verify that the DI-RVCSRN network proposed in this paper has good generalization ability in terms of performance metrics.Finally,based on the object detection algorithm Yolov5 n,a comparison experiment is designed based on the self-built blackboard dataset,and the performance metrics of the super-resolution algorithm DI-RVCSRN and other reconstruction methods that meet real-time reasoning are compared to verify the effectiveness of the reconstruction network DI-RVCSRN proposed in this paper.