This invention belongs to the field of aircraft
assembly technology, specifically disclosing a method for measuring wing and
fuselage assembly deformation based on a neural
network model. The method includes: acquiring image data from multiple different
assembly areas using
digital image correlation; preprocessing,
quality assessment, and simplification of the image data; expanding the dataset based on the preprocessed image data; constructing customized neural network models for different assembly areas and training the models using image data from the corresponding areas; deploying the trained neural network models to
edge computing nodes; and acquiring image data from different assembly areas in real time to obtain deformation data. This invention significantly improves the accuracy, real-time performance, and
system reliability of deformation measurement by adapting the customized network to the deformation characteristics of different areas and enhancing anomaly recognition capabilities through mixed datasets, thus meeting the industrial needs of online measurement for wing and
fuselage assembly.