Research On Visual Inspection And Recognition Methods For Foreign Objects In Glucose Liquid

With the improvement of people's living standards and the popularization of medical and health knowledge, the conception of medicine safety is gradually realized by more and more people. People have paid more attention to medicine quality. For a better development and existence, medicine production enterprise must continuously improve product quality, and apply modem science and technology to the production and processing of medicine. Currently, manual inspection is the popular method among most domestic enterprises because of the technological monopoly and blockade by foreign company, it is urgent to develop the automatic online equipments for product quality control based on machine vision in a glucose liquid product line, and it is a most important and difficulty problem for foreign objects visual inspection and recognition in glucose liquid in those equipments.First, this paper introduces the background and significance of related researches, summarizes the research and application status of machine vision technology in pharmaceutical manufacturing industry both at home and abroad, and then do further research and exploration on key technologies, likely image noise suppression, foreign objects inspection and recognition of glucose liquid visual inspection.Secondly, some noise model based on the obtained image of glucose liquid by system is analyzed in this paper, we research and implement image noise suppression using improved adaptive median filter and wavelet-based methods, the experimental results show that the methods in this article are good way to suppress image noise of glucose liquid, it also effectively reduces the loss of image valid information, and those methods laid the foundation for future foreign objects accurate recognition.According to the characteristics of foreign objects in glucose liquid, in this thesis, an foreign objects inspection method for glucose liquid based on Adaptive Mixture Gaussian Model is proposed and implemented. Adaptive Gaussian Mixture Models are applied to get region-of-interest (ROI) of the original glucose liquid images. Furthermore, the mathematical morphology operator is used to target segment to complete the inspection. Then a foreign objects recognition method for glucose liquid based on fuzzy support vector machines is proposed, this paper illustrates the algorithm ideas, and analyzes feature extraction for foreign objects and the classifier train method based on fuzzy support vector machines mostly. Experimental results demonstrate that the design of methods can accurately detect and recognize foreign objects in glucose liquid, and it can well satisfy the requirement in the glucose liquid product line.Finally, according to the requirements of the foreign substances inspection in glucose liquid product line, a suite of automated online visual inspection and recognition system based on machine vision in a glucose liquid product line is developed in this paper. We presented the main structure of this system, introduced the architecture and various parts of the constitution of the system hardware and software, and ways to achieve is described. Through the development of the actual system, it provides guideline for algorithm in this paper, and experimental platform for the practical applications of the algorithm.