Quality control in syringes production with image processing techniques

Quality control is an inseparable part of the production process both in handmade production and in mass production with machines. The quality control process is the process in which error rates that require more attention from production should be minimized. It is difficult and error-prone to carry out quality control by the human being, as with many jobs, due to fatigue, speed and / or imperceptible defects. Adding to these difficulties that the material to be controlled is made of transparent material such as an syringe, the error in the control process increases significantly. For this reason, it should be preferred that the quality control process in this process should be entirely with the machines. If it is not entirely by machines, it should be preferred that it is carried out by humans at least with machine support. The use of image processing in the human eye quality control process facilitates the development process of the system and increases its usability after design. In this thesis, a system was designed to assist the personnel performing the quality control process in the syringe packaging stage of an enterprise producing syringe using image processing techniques. In the literature review, no quality control study for the packaging stage of injector production was found. Therefore, the study is considered to be the first. In the study, a platform was created with the lighting system and a 12.2 MP industrial camera was used for image acquisition. Image processing techniques for solving problems; template matching, haar-cascade classifier, color filtering and morphological processing methods are combined. During the syringe and needle tip packaging phase, the presence/absence/quantity/foreign matter detection-control for syringe/needle tip objects was studied in the package image. Approaches used, purposes and accuracy rates can be listed as follows; object detection by template matching was 99.87%, object detection by haar-cascade classifier was 90.49%, object detection by color filtering was 90.85% and foreign matter detection by stain detection was 85.58%. Average working time of the methods; template matching was 2.13sec, haar-cascade 0.015sec, and color filtering-foreign matter search was 0.06sec. Keywords: Color filtering, haar-cascade, image processing, syringe production, template matching