Vision based process perception for smart manufacturing
Loading...
Date
2024-03-01
Authors
Μπασαμάκης, Φώτιος Παναγιώτης
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The recent growth in the field of robotics, coupled with the growing integration of cutting-edge technologies, such as Artificial Intelligence and specifically Deep Learning within the manufacturing industry, has enabled the development of flexible robotic systems with the ability to understand their environment and execute a range of tasks autonomously. This thesis introduces an innovative vision system tailored for various manufacturing operations, aimed at effectively addressing the challenges of object pose estimation and quality inspection via the utilization of state-of-the-art computer vision techniques coupled with Deep Learning Networks. Specifically, the system employs two distinct models, one for semantic segmentation and another for direct 6D pose estimation, in order to accurately estimate the pose of two distinct categories of objects. These categories include small components with similar geometries that require high precision in pose estimation and more substantial components characterized by varying geometric features. Moreover, the system incorporates advanced quality inspection capabilities by leveraging a deep object detection model to correctly identify missing or mispositioned components within assembled products. The proposed framework is applied and validated on a use case study inspired by the consumer electronics industry and is mainly focused on handling two different products: a trimmer head and a monitor cover. However, as discussed throughout the thesis, the system has the potential to be utilized for addressing similar scenarios without necessitating further adjustments.
Description
Keywords
Artificial intelligence, Computer vision, Machine learning, Robotics, Image processing