Research And Implementation Of Real-time Recognition Of Digitalinfomation In The Video Stream Based On Image Processing Technology

Unmanned aerial vehicles search for the target in some reconnaissance systems, thetarget coordinates information subtitles superimposed on real-time video, spread to theprocessing system, which intercepts related screen according to the need, inputs therelevant information to form intelligence. Read and re-entry from the video imagemanually, increase intelligence entry time and error probability, and lead to lowerreal-time intelligence and reliability since it can not obtain the target coordinatesinformation automatically,.The paper researches the video of the target coordinates information, extractescoordinate information from subtitles with each frame image of the video, and takespreprocessing method, as the gray scale on the image, image enhancement,binarization, gradient sharpening, discrete noise, image segmentation, and so on. Thepaper achieves coordinates rapid extraction by calling the mature OCR recognitionmodule. Specific research work have two parts in the following:The first section introduces the related theoretical knowledge of the DirectShow Filter, FilterGraph Manager, media types, VMR, DMO, etc.. Summary introduces the basic steps and methods ofDirectShow application program development.The second part introduces extracting coordinate information from each frameimage of video caption:(1) The part introduces the theoretical knowledge of the image preprocessing,including color model, histogram, image noise smoothing filter, the gray-scaletransformation of the image and image binarization, etc., The summary introducebinarization method of text image in the complex background.(2) The part studies extraction technology of subtitle text for the video, extractionstep of subtitle text for the video, method of subtitle positioning in continuous video andcharacter extraction method,method of the difference amplification and mixed-thresholdbinarization importantly, and the polarity judgment of the text and filter technology ofthe background pixel based on the connected domain analysis. (3) The part researches video capture API and API secondary development interfacedesign of character recognition. It chooses secondary development interface of C++dynamic link library in order to providing functions of the image frame with imagetitles from the active video and extracting coordinate information for users. This paperupgrades acquisition module of some reconnaissance system intelligence and achieve aseamless integration of legacy systems on the based of DLL interface.