Bearing Defect Detection Based on Fused Image Enhancement and an Improved Object Detection Model

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Mechanical compression and chemical corrosion during bearing dust cover production induce low-visibility pitting defects that traditional vision methods struggle to detect. This Letter proposes an improved YOLOv11 algorithm integrating dark-field illumination-based photometric stereo imaging and an adaptive segmentation pipeline to synthesize normal-map grayscale images. ESP2F, C3k2E, and SAUM improve multi-scale features, enhance defect-related channels, and restore weak defect edges and local textures, respectively. A cascaded CESFC and Spec-YOLO architecture filters normal samples to reduce inference overhead, outperforming baseline YOLOv11 in Precision, Recall, mAP@50, and mAP@50–95.