Automatic Detection of Sorbite Content in High Carbon Steel Wire Rod

This paper presents a method for the automatic detection of sorbite content in high-carbon steel wire rods. A semantic segmentation model based on DeepLabv3+ is established. The structure segmented, and prediction results are analyzed counted metallographic images rods marked manually. For problem sample imbalance, loss function Dice + focal used, perturbation processing training data added. show that this can realize statistics content. average pixel accuracy as high 94.28%, absolute error only 4.17%. composite application enhancement significantly improve robust performance model. In method, single image takes 10 s, which 99% faster than min using manual cut-off method. On premise ensuring accuracy, efficiency improved labor intensity reduced.