Research And Implementation Of Tobacco Laser Code Automatic Identification Systemwhich Base On Android

With the rapid development of the image processing technology, the technique of image pre-processing, image intelligent matching, and feature extraction have achieved great success in recent years. Nowadays the image progressing technology has been widely applied in various fields, for istance, in image generation, image translation and communication, robot vision, office automation and so forth. The development of image recognition technology has provided theoretical foundation for our study.In our study, the method recognizing cigarette’s laser code automatically is developed. The laser marked on cigarette case contains important information of the cigarette, and it is very important to citarette’s management. However, in traditional ways, to recognize cigarette’s authenticity and uncover the information of the cigarette, cigarette managers have to enter laser characters manually to recognize them by system, that is low efficienent and inconvenient.In this thesis, we develop a system based on Android mobile system, which help cigarette managers to recognize laser markings automatically and conveniently. In our study, the following problems are solved: first, how to catch the feature on images with complex background. By the reason of this light is not evenly or there are some pictures on cigarette cases, so it case the problem that how to catch the coverage area under this circumstances. Second, how to recognize chars from iamge accurately. Since using traditional algorithms directly cannot obtain high accuracy, sevearal existing algorithms are combined to recognize the laser markings. Third, how to build the program based on Android mobile system. For our study, the local threshold method is applied to threshold images, and the splite char method base on location and pixes is used to splite images and deep learning algorithms is employeed to recognize the chars. As a result, our threshold method is better than traditional method, and we can get a better splite result than other methods. The test result shows thst our system’s accuracy can achieve ninety percent on character recognition.