Research And Implementation Of Video Caption Recognition System Based On OpenCV

Video captain recognition is the technology that recognize image textin the video as pure text via computer, and is an important area of moderecognition applications. The video captain recognition technology isrelatively mature in developed countries due to the better developmentin computer vision and related industrial. However, the captionrecognition technology in domestic developed later, so the currentsubtitle recognition software is only useful for the image or is notsuitable for video with complex background. Therefore, the research ofcaption recognition system in video with complex backgrounds willbenefit internet video retrieval. Then the retrieval speed of video anduser satisfaction will increase.For this situation, the video caption recognition system based onOpenCV is researched. OpenCV issued by Intel company is a cross-platformcomputer vision library. It can operate on Linux, Windows and Mac OSsystems, also provides Python, Ruby, Matlab and other language interfaces.Therefore it becomes the preferred development framework due to therealization of image display and processing. C++language is used todevelop this system in the Visual Studio2008platform. The system hasa high accuracy and speed of recognition. Meanwhile the caption-framedetection, caption recognition, results storage and other tasks can bedone automatically. Thus the speed and accuracy of network video retrievalare greatly improved.This paper processed in strict accordance with caption recognitionto organize, analyze and function implementation. The first chapter introduced the research status of the video caption recognition systemand the main contents of research. The second chapter described thecurrent caption-frame detection technology. On the basis of the existingmethods, a new caption detection algorithm was proposed. The third chapteranalyzed and compared the various caption extraction algorithm. Thefourth chapter introduced the segmentation algorithm of character andthe normalization methods of character at first. Then considered anexisting deficiencies of recognition method based on feature extraction,a caption recognition method with feature fusion was proposed. The fifthchapter summed up and sorted out the key techniques and algorithms ofsystem implementation. Through the caption recognition of online video,the validity of the recognition system was confirmed. Finally, the papersummarized the research work and the application significance andprospects of the system were discussed.Successful development of this system can improve the accuracy ofvideo caption recognition, so the labour, material and financialresources in network video retrieval were saved a lot. Therefore, thedesign and implementation of this system is of great theoreticalsignificance and practical value.