Method and device for measuring average grain size of metal

The invention provides a metal average grain size measurement method and device, and belongs to the field of metal grain size measurement. Comprising the steps of predicting grain boundary information of a to-be-measured image based on a pre-trained grain boundary segmentation neural network model; wherein the grain boundary segmentation neural network comprises a high-resolution feature extraction module, a grain boundary center feature extraction module and a grain boundary center guided feature re-calibration module; the grain boundary center guided feature re-calibration module CFRM comprises a center feature guided channel feature re-calibration CCR and a spatial feature re-calibration CSR, and the grain boundary center guided feature re-calibration module carries out center feature guided channel feature re-calibration and spatial feature re-calibration based on the grain boundary center feature extraction module; and calculating the average grain size grade based on the product of the grain boundary in