APPLICATION OF IMAGE PROCESSING ALGORITHMS FOR THE AUTOMATIC ASSESSMENT OF WORK PIECE QUALITY

The Industry 4.0 or Fourth Industrial Revolution is causing a change of conventional manufacturing and mechanical practices, by adding smart IT innovation. This fundamentally centers around the utilization of large scale machine to machine communication (M2M) and Internet of Things (IoT) arrangements to give expanded automation, improved communication and self-observing. Particularly, in this context, smart machines are developed, which  can investigate and analyze issues without the requirement for human mediation. Casting products are important components and have a crucial role in large equipment. Casting is a manufacturing process in which defects may form while pouring the metal. Surface defects (e.g., blow holes, shrinkage, and metallurgical defects) are common in casting and they can be detected by visual inspection. However, since this is a time consuming process, in industries with large production the quality control can be a crucial factor. Machine learning algorithms can be an aid to this problem by allowing time saving and neglecting human errors.  The aim of this thesis is to apply image processing and machine learning algorithms for the automatic assessment of the quality of a casting. A public dataset of casting images has been used as a case study. Three image processing techniques have been used (resized data, adaptive equalized data and HOG data), and different classification algorithms have been tested and compared.