Research On Recognition Character In Express List By Mobile Phone Camera Based On The OpenCV

With the rapid development of e-commerce, the development of logistics industry presents unprecedented pomp. The staff need to put a lot of express single information into the mobile phone every day, it has an important practical application value to research a kind of method which can automatically store the information in the form of text to mobile phones. The existing OCR products are mostly based on PC platform. With the popularity of handheld terminal products, especially the widespread use of smart phones, character recognition and its application in mobile phone have become one of the research hotspots and difficulties. Upon all the above content, We can express list image by mobile phone photo and the express list of important information on the form of a list text is stored to mobile phones through the programming process.This paper studies the technology of image preprocessing, which improves the existing linear blanking method by increasing the blanking of vertical line on the basis of blanking horizontal line. Based on edge processing, it designs a combinatorial positioning algorithm combining image morphology processing, contour detection and external rectangular positioning. This method conforms to the demand for text positioning in express single. Character segmentation processing in digital way, analyzes the special relationship between digital's inner and outer contour, and is designed as a fast digital segmentation method based on inner and outer contour detection, this method works very well when dealing with Split Fail problems caused by overlap between characters, however these problems cannot be solved by vertical projection segmentation algorithm. In the stage of Chinese characters segmentation, Chinese character segmentation method based on connected domain and vertical projection is discussed, after which we add the wavelet transform method. Besides, we analyze the structures commonly used in character recognition, and then choose features such as digital intersection of outline, Hu moment feature extraction and convex defect as characters to extract; For Chinese characters, we choose the coarse grid features, the density of Chinese characters strokes, the corner feature. In the stage of recognition,we adopt Euclidean distance and KNN neighbor classifier together to recognize characters.Lastly, we develop our own programming module on the Android mobile phone platform using the Opencv library for core programming as well as loading OpenCV-2.4.6-android-sdk based on integration environment:adt-bundle-windows-x86. As a result, our method can Preliminary recognize express single words. Our Experiments prove that, this research has a certain theoretical and applied value in the field of text recognition.