研究目的
To address the issue of insufficient training data in the visual measurement of surface roughness by proposing a novel method based on inductive transfer learning to establish an advanced roughness predictive model.
研究成果
The proposed method significantly improves the accuracy of surface roughness measurement with insufficient training data, achieving an average relative error of 12.57%. It demonstrates the feasibility of using inductive transfer learning and simulated data to enhance model performance in visual roughness measurement.
研究不足
The method's performance is contingent on the quality of simulated images and the similarity of aliasing effects between actual and simulated domains. Potential areas for optimization include the threshold selection for the ARA index and the generalization of the model to different surface types.
