Research On Adhesion Anti-fake Code Recognition And Implementation Of Handheld System

Character recognition is a branch of computer vision which is widely researched and used in recent years.Character recognition has been widely used in automobile license plate recognition,identification systems,character codes recognition and so on.At present a lot of businesses print anti-counterfeiting code on the packaging to implement anti-fake.Recognition system to automatically recognize the anti-fake code and compare it with the database to identify the authenticity of goods,has high research value and broad application prospects.However,most existing mature technology are applied in the cases of simple background,regular character,and PC or server platform,and can't be directly applied in some certain scenes.The goal of this paper is to design a recognition algorithm and its handheld system to recognize the anti-fake code on cigarette packaging which is in complex background,adhesive,and has low contrast.The main contents of this paper are as follows.(1)Image enhancement,binarization,denoising and other kind of preprocessing is implemented.Some mature binarization algorithms are implemented and compared in the key part of preprocessing: binarization.To solve problems such as complex background,uneven illumination,we use local binarization algorithm to remove background,edge detection to extract the character outline,OSTU algorithm to roughly determine areas for stroke.Eventually an algorithm which is a combination of OSTU algorithm,Sauvola algorithm,Canny edge detection algorithm and background estimation is proposed and finally a good effect of binarization is achieve.(2)Comparison of several common segmentation algorithms is made and an improved algorithm is proposed to segment adhesion characters.In this algorithm,the top and bottom outline are obtained by vertical projection,and then all the extreme value in the histograms are found as alternatively cut points,and finally,the estimated height,width,the number of characters and the other characteristics are took into account to find the best combination of cut points from the extremum value sequence.In the final recognition phase,a large amount of sample images are collected and an appropriate sample library is gained and then the training and recognition process are implemented based on the ANN method to achieve a Recognition rate of 95%.(3)Samsung S5PV210 is selected as the processor and the hardware circuit of the handheld identification system is designed.The hardware consists of core board and peripheral circuit board.The core board includes S5PV210 processor,NAND Flash,DDR2 memory,power supply,clock and reset circuit.And the peripheral board includes power management circuit,touch and display control circuit,storage circuit,USB extension circuit,wireless communication interface circuit,image acquisition unit and so on.Android is selected as the operating system and the anti-fake code recognition algorithm and the application software are implemented on it.