Software Design Of Video OSD Automatic Verification System

With the development of network multimedia technology, digital video system is widely used in banking, transportation, public order, home appliances, handheld devices and other fields. Before deploying these monitoring systems or using video equipment, the correctness of OSD (On Screen Display) display must be assured carefully. However, traditional verification tasks are throughout completed by human labor, the way of testing which consumes lots of material resources as well as human resources. As a result, the artificial verification is rather low in efficiency. Therefore, how to quickly and accurately verify OSD is an urgent issue to improve the productivity of video system.To deal with low efficiency and long-time costing in verifying OSD. this thesis devised an automatic OSD verification system which can automatically locate and verify OSD letters in a picture. According to the high recognition rate and low time-consuming, this system could supersede traditional labor verification, which has a strong engineering research value. This system is capable for intercepting one frame from video stream, first pre-processing technology is adopted to improve signal-to-noise ratio, then four statistical characteristics of OSD character, which are vertical projection, letter block scale, field edge value as well as left and right border extremum, is synthesized to locate OSD; afterwards, the character projection information and the priori character information are calculated to accurately segment character one by one, followed by binarization with improved Otsu algorithm, Gabor features and Improved BP neural network is finally used to recognize and verify these characters by comparing with string in database. Experiments show that this system costs average97ms per recognition of one frame with a correct rate at92.7%, totally accomplishing rapid and efficient requirements.