A Machine-Vision-based Detection And Counting System For The Packing Of Cigarette

How to survive in the drastic competition is a serious problem which should be took in to account by the companies. As a result, the reliable and efficient detecting of the product's quality is necessary. The online quality detecting system based on the computer vision, which can pick out various flaws of the product swiftly and correctly, increases the product's eligibility, promotes the core competition ability of the company and shows its vast applying prospective.First, this paper brings out the research background, actuality and developing direction of our task. The common problems found in the current online detecting method are analyzed.Then, after discussing the tech requirements and characteristic of the online detecting system based on computer vision, the designing scheme of software and hardware are narrated.According to the character of the image of the packing of cigarette, necessary pre-disposing is needed first to improve the image's quality. For the sake of contrasting the model image and the actual image, which is critical to the flaw detecting, the two should be matched. This paper proposed a swift image matching algorithm based on the sketch, which can bring efficient matching results.After the detecting, it is necessary to count the number of packing of cigarette. A new counting algorithm based on mathematical morphology has been put forward in this paper. The algorithm can highly enhancing the veracity and meeting the shortfall of the mechanical sheet numbering equipment which widely used in our production and life.In the development implementation, a practical on-line quality detection and counting system for the packing of cigarette is presented, and the proper performance has also been demonstrated in this thesis.