Redundant object detection method for civil aircraft assembly based on machine vision and smart glasses

Abstract

Abstract Slow detection of redundant objects and low accuracy in assembly lines, particularly in the setting of civil aircraft assembly, are tough and challenging problems. To address these issues, a redundant object detection method based on computer vision and augmented reality (AR) smart glasses is proposed in this paper. The method uses AR glasses as the image collection hardware and takes the live image collected by the camera as the input of the proposed deep learning machine vision model. The proposed model, the Feature Pyramid Networks-CenterNet, is inspired by CenterNet and combined with multi-scale feature fusion to solve the problem of low detection accuracy of small-scale redundant targets. The weight factor of the loss function was set according to the proportion of small targets in the dataset, which solves the problem of an unbalanced proportion of large and small targets in the training samples. The proposed network model was validated on the PASCAL Visual Object Classes public dataset and the self-built redundant object dataset. The results showed that the new method can detect seven redundant objects with a mean accuracy of 74.49% within the visible range of smart glasses within 200 ms. The research provides a new reference for the quality process management of civil aircraft assembly.