Research On The Character Detection And Recognition System Of Beer Bottles Based On Machine Vision

In recent years, malignant incidents take place occasionally causing by beer bottle explodes. These incidents often lead to deformity even death. One of the reasons that lead the fact is the bottle been recycled too much time to exceed the time limit. One of the other reasons is the abuse of some non-special-beerbottle.Eliminate unqualified bottles and non-special-beerbottles is usually handwork, but the workload and the cost are dissatisfactory. It is very meaningful to make the computer finishing this work automatically.Study goal is to make computer extract and recognize the characters of"B"and the marker of year from the image automatically. If mark "B" is exist and service time is proved within two years. That indicates the bottle is qualified, otherwise unqualified bottle.The whole system includes the imaging system and software part. The software part includes pretreatment, characters segmentation and characters recognition. This paper presents a binarization algorithm with many-thresholds-value, and a segmentation algorithm basing on projection of gray image taking character width, character shape and the interval width into consider. Finally proposes model match with different resolution and configuration characteristic algorithm to recognize the character.A beer bottle characters recognition system has been developed based on the above algorithm, with software in Visual C++. We preformed experiments with a sample of large numbers of recycle beer bottles. In these experiments, 95% beer bottles were recognized. In conclusion, the algorithm presented in this paper with high stability is fast in calculation and effective. Hence, this system may apply to the industrial environment.