| The defect detection and size measurement of medical bottle is an important element in the inspection of tube-type bottle. Traditional artificial visual method has several shortcomings, such as low efficiency, high rate of undetected and bad accuracy. The purpose of this thesis is designed to built an inspection system of medical bottles based upon machine vision. In this thesis, in allusion to the image characteristics of medicinal bottle, we design and test the following aspects:Firstly, We designed an overall scheme for the system, including the selection of acquisition equipment, light source and lighting design. Light source and lighting design solved the problem of bottle light reflection and the problem of heterogeneity. Hence, the system can acquire better quality images to facilitate the follow-up testing of image processing.Secondly, By analysis and comparison of various image processing algorithms, we design a set of new image processing algorithms in accordance with the characteristics of the target in this thesis. The algorithms are composed of using a improved median filter method for image smoothing and detecting edge by combining the OSTU image segmentation method and morphological edge detection. A series of experiments are conducted to verify the feasibility. Moreover, the result is very competitive. The algorithms greatly facilitate the following defect and size detection.Thirdly, According to the specifications of appearance and size of bottles, we make the defect detection and size parameter measurement to the bottle. We proposed the edge pixels scanning method for bottle defect detection. Moreover, we employ the improved HOUGH transform method to implement the center position location and size measurements. Finally, we make error analysis for the experimental results.Finally, Application interfaces are designed. In allusion to different kinds of users, we design the main interface and sub-interface for production and engineers, respectively. |
PDF Full Text Request