Abstract:For surface roughness measurement methods based on machine vision,prediction models are mainly established through image feature information and roughness correlation indicators,but the lack of sample size often makes it difficult to train an effective model,resulting in low measurement accuracy.In view of the above problems,a small sample grinding surface roughness measurement method was proposed.The image acquisition system was established to collect images of grinding surfaces with different roughness levels as the original samples.The virtual sample generation algorithm was used to expand the sample size.The gray level co-occurrence matrix was used to extract the sample texture features.Finally,the prediction model was established through the neural network.The experimental results show that after the sample size is expanded,the accuracy of surface roughness measurement increases from 80.4% to 97.2%,which proves the feasibility of this method and provides a theoretical basis for the on-machine inspection of small sample grinding surface roughness.