A novel surface roughness measurement method based on the red and green aliasing effect

Abstract

Abstract A common requirement in many machine learning algorithms is that the training data is sufficient. However, in the visual measurement of surface roughness, this requirement often can't be meet due to the reason that it is time-consuming and expensive to process and label the training samples. To address this issue, this paper proposes a novel method to establish an advanced roughness predictive model with less standard training samples based on inductive transfer learning. The experimental results show that the proposed method has superior measurement performance, and can maintain the average relative error of 12.57% even when the training data is insufficient. This indicates that the proposed method can provide a new strategy for improving the visual roughness measurement performance.