Design And Implementation Of Video Streaming Text Recognition System Based On Distributed Scheduling

In the current,some office systems need to integrate with other systems,but because of the various systems of some technology and providers are different,so it is difficult to integrate.With the development of streaming media technology,the problem of data fusion of multiple systems can be achieved by capturing the fixed position of the office system.But because the office system may has a lot of users in the same time,so we must solve the problem of character recognition under concurrent real-time video streaming at first,but if the system encounter some exceptions,we hope these is another solution which can process the stored video files in the late.In order to realize the real-time video streaming process,two problems need to be solved.At first,for the video streaming process,the processing speed of video streaming and video files is different,if not timely process video streaming will lost some frames of video which will affect the image quality of the decoded.So in the video streaming decode,with FFmpeg library,we use thread pipelining to realize the reduction of the loss of frames in the video streaming process.In decoding and OCR recognition in the video,because of the existence of relatively large differences in processing speed.At first we separate the decoding and character recognition module,thus it can reduce the coupling function and code.To solve the mismatch problem between decoding and OCR recognition speed,Redis is used as the buffer storage between the two.Redis is a database work in memory,it not only has the general characteristics of NoSQL database,but also it has very fast speed to access the data.In order to solve the problem of process a lot of real-time video streaming in the same time,a distributed task scheduling framework named Gearman has used.In addition to dealing with real-time video stream,this paper also proposes a method for the identification of video files.The method is to upload the video file to the specified streaming media server,so as to realize the recognition of the video file.On the one hand,the use of distributed systems to achieve the video stream decoding in a large concurrent real-time video stream processing;the other hand,through a upload client to achieve the conversion of video files into video streams.So that the distributed video stream processing system can not only deal with real-time video streams but also can process video files.