Research On The Key Technologies Of Laser Reproduced Chinese Character Recognition

Laser reproduced Chinese character recognition is the key technology of laserholographic encryption technology. It provides a machine-readable way to identify theauthenticity of laser holographic anti-counterfeit labels, greatly improving the security.Therefore, research on the anti-counterfeiting machine recognition has importantsignificance and practical value.Focusing on the crucial technologies of laser reproduced Chinese characterrecognition, a deal of research and experimental analysis is stated in this thesis, includingthe preprocessing techniques, liner feature detection techniques and the character imagerecognition technology. According to the particularity and the complexity of the originalimage, a variety of image preprocessing methods are combined to be used. Firstly, theimage’s R,G,B components are separately conducted a top-hat operation. Then the threeresults are added together to become a grayscale image. From the grayscale image, theOtsu’s method can be used to create a initial binary image. Afterwards, an open operationis conducted on the binary image, and its result will be the seed points for the regiongrowing of the grayscale image, eventually an ideal binary image comes out. Second,Canny operator is used to extract the contour information of the Chinese characters, andthen Beamlet transform which is also innovatively used for the liner feature detection ofthe contour removes some false edges. Therefore, a more accurate character’s profile canbe obtained, and is ready for the recognition of Chinese character. Finally, by making fulluse of the Chinese characters’ profile information, a method based on calculating themaximum correlation coefficient of the character’s profile and the template’s outline isselected, to get the recognition result for each character.Experiments validate that for the original images collected at different times andunder different lighting conditions, the algorithm of each step can get a satisfactory result,no matter at the preprocessing, extracting the liner feature or Chinese characterrecognition. In addition, it is also verified that different scales of Beamlet transform has astrong point of extracting the thin line in very noisy image, and this advantage can beapplied to a wider range of areas as well.