DLCS-YOLO Model for Detecting Defects in Long-Distance Oil and Gas Pipelines

Area of Science:

• Artificial Intelligence
• Computer Vision
• Materials Science

Background:

• Permanent magnetic field perturbation (PMFP) is crucial for oil and gas pipeline defect detection.
• Existing methods struggle with high false-positive rates, noise interference, small defects, and multiscale variations.

Purpose of the Study:

• To develop an advanced YOLO-based model for improved PMFP-based pipeline defect detection.
• To enhance accuracy, reduce false positives, and improve adaptability to various defect types and scales.

Main Methods:

• Introduced the Deformable Large-kernel Context-fused Spatial (DLCS)-YOLO model, an enhancement of YOLO v11n.
• Integrated C3k2-Deformable Attention and SPPF-LSKA modules for robust feature extraction.
• Employed Context-Guided FPN and SEAM for superior feature fusion and detection adaptability.

Main Results:

• Achieved mAP@50 of 94.5% and mAP@50:95 of 64.7% on a custom dataset with 6.2 GFLOPs.
• Demonstrated a 3.1% increase in precision and improved mAP values compared to the baseline YOLOv11n.
• Showcased enhanced robustness to noise and superior performance in identifying small-scale and multiscale defects.

Conclusions:

• The DLCS-YOLO model significantly improves the accuracy and efficiency of pipeline defect detection.
• The proposed model offers practical value and broad application prospects for oil and gas pipeline integrity.
• Advanced attention mechanisms and feature fusion strategies are key to overcoming current detection challenges.