The Design And Implementation Of Image Recognition System Of Cigarettes

In order to conform to the enterprise development trend, whose center is modernization and informationization, many domestic tobacco companies have adopted modern logistics management method. The traditional manual operation was replaced by the automation of distribution system of cigarettes. Automatic cigarette selection is an important sub-link of distribution system of cigarettes. The image recognition system of cigarettes is used in the process of sorting to improve the efficiency of sorting and reduce the labor intensity. The system can distinguish the brand and type of the packaged carton of cigarettes, and can accomplish the automatic review function through the computer.This Paper with the whole working flow of the image recognition system of cigarettes for the main line carries on the organization. The introduction Part introduces the background knowledge and the research meaning of the system,design of the image acquisition device. Then, analyse system requirements and the characteristic of cigarette image, finish the overall design of system software, and determine each part of the algorithm research direction. Next, analyse some idiographic algorithm of edge detection,segmentation,feature extraction concretely. Based on the color and size of cigarette images, formulate solutions about elementary screening and fix on a decision-making plan of recognition.There is a multi-level image preprocessing algorithm, called "The outer boundary sampling method", in this paper. The outer boundary sampling method, which is targeted, makes use of color,shape,size and location of cigarette images. The image preprocessing is divided into several levels by this method. Each level can solve different questions, and effective images used for the recognition can be obtained by synthesizing the results of several levels. In the feature extraction, the method combines principal component analysis and Fisher linear discrimination, which can get the feature vector of image with conversion matrix. The recognition decision is based on the feature vector.At the end of the paper, enumerate experimental results of several parts of the system, summarize research results of the topic, discussed the developing direction on this topic.