Method and Application of Spraying Developer Defect Detection in Three-Dimensional Optical Scanning Measurement Process Based on YOLO v7

Abstract

In industrial production, three-dimensional scanning and inspection technology is widely applied, with spraying developer playing a critical role in this process. During the application of spraying developer, the spraying defects generated on the surface of workpieces can greatly affect the accuracy and integrity of the three-dimensional scanning data construction for the products. Currently, the detection of these spraying defects heavily relies on manual visual observation, a process that is both time-consuming and labor-intensive. Moreover, there is no standardized approach, leading to uncertainty in the subsequent construction of three-dimensional scanning data. To address this issue, this study proposes a deep learning algorithm based on YOLO v7 target detection. The research focuses on commonly used inspection parts in the field of three-dimensional optical scanning measurement, constructing a dataset specifically for spraying defects. Real-time target detection is then conducted on the workpieces within an actual production environment. Through model evaluation, it was found that the neural network in this study achieved a remarkable accuracy of 0.996 on the test set, with the highest confidence level of 0.96 during real-time inspection of the workpieces. These results demonstrate the superior accuracy of the proposed method in detecting defect targets, thereby offering valuable insights for evaluating spraying defects and enhancing the quality of three-dimensional optical scanning data construction.

Keywords

• Three-dimensional scanning
• spraying developer defects
• deep learning
• computer vision
• image detection