Automatic Detection Of Foreign Matter In Bottle Based On Image Processing

With the development of society, people pay more attention on food safety issues, the problem of the quality of the beer is becoming more and more important to the consumers and producers. For example, insoluble foreign matter inside the bottle is appeared, bottle is not perfect and so on.For such problems, traditional artificial light inspection method of liquid products can not meet the requirements of producers and consumers. On the basis, this thesis designs an automatic detection system for foreign bodies in the bottle based on image processing, to analysis and identify the foreign bodies in the bottle.On the one hand, this thesis introduces the background and significance of the thesis, the machine vision's application field and the current situation. On the other hand, starting from the actual production line of beer, a algorithm of machine vision is used to design and develop a set of test system. The system mainly includes a lighting system, a testing equipment and a processing software. Finally, the scheme of the detection system and the working flow are given.The target detection is the key point of this thesis. In the image, we analyze the target in a effective dectection area which is calibreted by the gray and geometric information. Effective area consist two parts:the body and the buttom of the bottle. Foreign bodies in the bottle body region include small particles and the bubbles at the liquid level. Foreign body in the bottom area of the bottle includes a large area at the bottom of the bottle and small particles.For the goal of small particles of foreign body, the bottom and the body of the bottle region's detection algorithms are the same:we adopt the first-order differential detection operator and mathematical morphology operation to obtain all the connected domains, then the characteristics of the connected domains are analyzed. According to the geometric and gray features we can initially identify the scope of small foreign objects. In this step, this thesis proposes a improved method for the detection of small particles, removing the interference of small foreign objects based on the characteristics of printing. Finally we can finish the detection of small foreign matter. For the bubble, the combination of the Histogram of Oriented Gradient and the Support Vector Machine is used to classify the bubbles and the jamming targets. The experiment shows that the method has a good classification results. For the large foreign objects sink in the bottom of the bottle,after removing the noise and suppressing the background, we adopt the OTSU method to segment the image. As the large area of foreign matter is connected with the bottom of the bottle, we depict the upper part of the curve on the bottom. Finally, we use PCA and Gabor wavelet to extract the characteristics of the bottom curve, then the characteristics were classified by SVM. Finally the accuracy of the two methods are compared and analyzed.This thesis complishes an automatic detection system for foreign bodies in the bottle based on image processing. The system not only improves the detection accuracy and speed, but also has a great economic benefits.Finally, the shortage of the experiment and the future work were discussed.