Research on Surface Quality Monitoring of Workpiece Based on CornerNet

In order to monitor the surface quality of a workpiece in the production line in real-time, a method based on CornerNet is proposed. First of all, this method uses the anchor free target detection framework, which avoids the problems of data imbalance and introduction of more hyperparameters caused by traditional deep learning models in the design of anchor. Secondly, the ResNet50-vd that incorporates the idea of deformable convolution is selected as the backbone network of CornerNet, and uses the feature pyramid structure to enhance the feature learning ability of the model for better real-time performance and robustness. A machining production line in Guizhou Province was used as an experimental platform to verify the proposed method's feasibility and effectiveness. Experimental results show that the method has various advantages such as simple and clear structure, strong adaptability and so on. In addition, the detection accuracy and speed meet the needs of industrial scenarios.