Deep convolutional neural network object net model based cognitive digital twin for trust in human–robot collaborative manufacturing

A Cognitive Digital Twin (CDT) is an artificial intelligence enhanced version of a digital twin, mirroring and learning from its physical counterpart. A cognitive digital twin in Human–Robot Collaborative (HRC) manufacturing generates synthetic robot data to build trust through simulating and monitoring collaborative behaviours, enhancing efficiency and safety. This encounter issues like restricted trust, inefficiency in extracting crucial details, difficulties in precise detection and alignment of bounding boxes with object boundaries, and high computational complexity. To overcome these challenges, a Deep Convolutional Neural Network Object Net (DCNNONet) model is proposed. The adaptive non local moment mean filter enhances image preprocessing, optimizing noise suppression and texture preservation. Utilizing proposed model, which includes featuring EfficientNetB7, a hybrid pixel unshuffled network, and long-edge decomposition rotated bounding box encoding, effectively tackles these issues, surpasses in feature extraction and precise object detection. Additionally, the introduced enhanced growth optimizer refines model parameters, reducing overall computational complexity. The introduced scheme exhibits better performance, attaining high accuracy, mean average precision, and sensitivity, and F1-Score of 99.96%, 98.5%, 99.75% and 99.89%, respectively. This model enhances trust by providing reliable insights, increasing accuracy in manufacturing scenario interpretation, and improving precision, optimizing performance in collaborative environments.