Design Of Handheld Character Recognizer Based On MCU

With the increasing development of image processing and pattern recognitiontechnology, people have done much re-search in character recognition, and more andmore relevant applications have been used in our life and work. The online recognitionsoftwares are relatively more in these identification products, the text manuscripts areinput into the computer rapidly as a kind of image through the scanner and the texts incolor images are recognized through the recognition software, which is not onlytroublesome operation but also not easy to move because the environment is relativelyfixed. Moreover, there are some handheld identification products, but which areexpensive and not easy to repair. In order to make the handheld character recognitionmachine convenient for people to carry, it has important practical significance todevelop handheld character recognition machine that is low cost, higher recognition rate,and good stability.Combined with the OCR(Optical Character Recognition) and technology ofembedded system development, this study has developed a simple, portable andpractical hand-held character recognition system using a low-cost, high-performancesingle chip microcomputer, which includes the collection, storage, identification, andother functions. In order to realize the system function, the related research work isillustrated in the following parts in this paper.①Refer to the related research of the domestic and foreign scholars in the field ofcharacter recognition, and according to the needs of practical application, theanticipated goal of the system was proposed, and the implementation method of thehardware of the system were given. The STM32F103ZET6made by ST was used as themain control chip.②Taking into account the development costs and prices, the design of thehardware unit and the soft implementation, this paper selected the OV7670camerasensor, and the AL422B FIFO flash-memory chips, the collecting module of the systemwas designed. The TFT-LCD that is2.8inch was used as display module, and whichis controlled and implemented by the unique FSMC (Flexible Static Memory Controller)memory expansion technology, the storage management of SD card was implementedby FatFS file management system. The three module task through by theimplementation of μC/OS-II system. ③For the recognition algorithm of character, in order to get the target informationeffectively, the image preprocessing techniques such as the grey scale processing,binarization, the bottom frame and the elimination of isolated point noise, andsegmentation of characters and so on. The binarization used the classical method ofOTSU, projection method was used to position the characters to be processed in thesegmentation part. Based on the characteristics of characters, this paper synthesizedstructural features and statistical characteristics of the characters for the feature pointextraction, and the first round of the feature points were extracted and encoded for thecharacters that have been segmented, the characters that had the same coding were putinto same subset, then the second round of feature points extraction were completed.The characters were effectively expressed through two rounds of coding.④Based on the construction of the handheld character recognition system, for thecommon62characters, the function and performance of the system were tested, and theexperiment results were analyzed.Many experiments show that this system can achieve requirements of the accuraterecognition character, and the digital recognition correct rate is95%, characterrecognition correct rate can reach92%. For the speed of identification, it canrecognition two characters per second and meet the requirement of the real-timeprediction, which provide a valuable reference for practical application.