The Research Of Pharmaceutical Packaging Inspection Technology Based On Machine Vision

As the high-speed development of technology and more detailed in producing process as well as the more complicated production, the manufactures are paying more attention to the inspection of product quality step by step now. Visual detection technology is increasingly being applied in various fields. And it can substitute the manual for fully automated product testing and certification process, or even the entire production process control. In pharmaceutical packaging, conventional quality inspection method by manual packaging is with high work load but low efficiency, and it can become error check and lacking check easily. Using a machine vision method can increase the inspection efficiency and success rate while ensuring the quality of production, and it can also alleviate the worker load and reduce the product cost. Firstly, this paper analyzes the international research situation and the need of machine vision. The specialties of machine vision are summarized, and the application of machine vision is generalized.Secondly, an automated visual inspection system based on an aluminum plastic blister packing product line is developed in this paper, which is composed of imaging unit, image processing unit and classifying inspection unit, using image processing and classifying inspection technique. This paper researches the core technique of imaging system in machine vision deeply, and a scheme of image grabbing based on image card and CCD camera is designed. An appropriate scheme of system lighting is described for complicated pipelining field in tablet packing, which is satisfied with request for clear plastic blister tablets'image.A method based on obtained plastic blister tablets'image by system is analyzed in this paper, which is applied the image processing and classifying inspection technique. Focus on inferior product in plastic blister tablets'packing, an inspection method based on Support Vector Machine is proposed. According to statistics learning theory, this paper analyzed the classifier train method based on Fourier Descriptor using for obtaining eigenvector and SVM using for classifying mostly. After comparing with the inspection method based on BLOB analysis and the inspection method based on template matching, the result of that proved the inspection method based SVM has a better inspection effect.At last, an image processing and classifying inspection arithmetic designed for plastic blister tablets is implemented by software, and it also describes the function of selected software development kit, the general framework of software and the flow of software.