On-line Detection Technology And Realization Of Bottle Packaging Based On Machine Vision

In the process of medical production,it is an indispensable link to detect the location of the packing label with reasonable position and clear content.At present the work is done mainly by manual way which spends a lot of manpower and resources.The result of label detection is not only high,but is easy to be affected by other factors.Therefore,this paper designs a detection system of label based on machine vision technology.First,according to the current status of machine vision technology and combined with the indicators and requirements of label detection,we fulfil the improvement of actual production line.From the point of view of the selection principle and parameter setting,the hardwares used in the paper are briefly explained.Then,the related algorithms of label detection technology are studied.With the help of traditional image processing technology in OPenCV lib,such as image grayscale,image denoising,image geometric transformation and image segmentation,the work of image preprocessing is completed.The location of label's edge is judged by projection result and the label bar code is detected by the Open Source toolkit ZBar.Through the Faster-R-CNN network based on deep learning,we finish the detection of fold and stain on the label.Finally,with the help of Qt Creator and C++ language,the overall integration of the system is finished.The simple GUI is realized with the help of Qt.Combined with the theoretical analysis and test result,the feasibility and practicability of label detection system based on machine vision is confirmed,which lays the foundation to the further medicine related detection.