Enhanced CNN-Based Object Detection with Safe Human-Robot Interaction for Quality Control in Food and Textile Industries

This thesis investigates the implementation of enhanced deep convolutional neural networks (CNNs) and safe human-robot collaborative interaction to detect objects from a variety of domains. The study uses two main methodologies: contamination detection on food products and textile fabric defects detection. In the first methodology, a deep CNN architecture is created and trained to detect contamination on food packaging, thereby guaranteeing that food quality regulations are followed. The model is built into a framework that enables safe interaction between humans and machine, reducing human efforts and lowering the risk of contaminations in food packaging. The second methodology focuses on textile defect detection, with a CNN-based system developed and enhanced to find defects in fabrics. The system detects many types of defects with adequate accuracy using modern deep learning techniques, contributing to quality control operations in the textile industry. Both techniques focus emphasis on ensuring safe collaboration between humans and machines, allowing for seamless contact and increased efficiency in industrial settings. The outcomes of this research have important impact on the development of intelligent technologies that can execute complicated tasks while prioritizing safety in human-robot collaboration contexts. Furthermore, this thesis includes two review papers that provide detailed insights into cutting-edge research and advances in human-robot interaction, contextualizing the study findings within the broader field of robotics and artificial intelligence.