A Method for Bearing Surface Defect Detection Based on Improved YOLOv10

In industrial manufacturing, bearings play a crucial role, and the accuracy of bearing defect detection is essential for the stability and safety of machinery. However, due to the complex environment and diversity of faults in actual working conditions, existing bearing defect detection methods are prone to missed or false detection. This study proposes an improved YOLOv10-based method and system for detecting surface defects in bearings, incorporating the EMAttention attention module, the SPPELAN module from YOLOv9, and replacing the Backbone with FasterNeT to construct an enhanced bearing defect detection model. The improved model increases detection accuracy by 1.3% while reducing computational load and improving detection efficiency.