The defect detection of mechanical parts is a key link in industrial production, but traditional manual detection methods are inefficient and difficult to meet the needs of modern manufacturing. To overcome this challenge, this paper proposes a mechanical part visual defect detection method based on deep vision sensing technology, which integrates Transformer and YOLOv5 models. First, the YOLOv5 model has improved its ability to extract part features. Subsequently, a new detection model was constructed by combining the pre-trained transformer structure with the improved YOLOv5. To validate the performance of the proposed method, we selected two real datasets for experimentation. First, we evaluated the improvement of key elements in the model and found that the improvement of certain elements can significantly improve the average detection accuracy. Next, we compared the proposed model with several typical detection methods, including SSD, Faster R-CNN and RetinaNet. The experimental results show that under the same dataset and measurement standards, the performance of this scheme is superior to the comparative methods, with higher detection accuracy and faster detection speed. The contribution of this paper lies in proposing an innovative visual defect detection method for mechanical parts, which combines the advantages of Transformer and YOLOv5 to achieve efficient and accurate detection of mechanical part defects. This method not only improves detection efficiency, but also provides strong technical support for quality control in industrial production.
Read full abstract