Development of vision-based recognition approaches for multiple anomaly detection in objects on conveyors

Machine vision is a system which obtains automatically data from images via utilizing image processing techniques in order to analyze a motion or an object. Nowadays, Quality Control and Recognition systems are widely using this system. Machine vision prevents human-induced problems such as inconsistency, eyestrain, illusion, etc. in these systems. In addition, these systems are working at a speed beyond the reach of the human eye and work tirelessly for 7 days 24 hours. Quality Control and Recognition systems realized with machine vision technology could achieve the control of any product in less than one second. Thus, they could control all products coming out of production in a very short time. In this work, a quality control and recognition system based on vision centric approach has been developed. In this system, images of the plastic market and plastic store bags were obtained from the plastic bag manufacturer Erhan Plastik factory, which is a local company of our city, for image sampling. A vision based quality control and recognition system has been developed in this direction since cutting and printing errors are occurred during the production of bags in plastic bag factory and an application is needed to separate these faulty bags from the bags carried on the conveyor system. All algorithms which are developed via utilizing image processing techniques in this work are original and new. In the vision-based quality control and recognition system, the bag images carried on the conveyor system were taken with the help of a camera and different methods which detect two different anomalies on multiple bag images were proposed. In application results of this work, accurate and fast anomaly detections in the cutting process of plastic market bag production and the printing process in plastic store production were confirmed with experimental images.