Abstract:Abstract Pipeline transportation serves as the primary method for conveying oil and gas, with welding being the predominant means of connecting pipelines. Nevertheless, weld defects frequently occur at circumferential welds due to various factors including the welding process and environmental conditions. Failures due to these welding defects threaten the safety of pipeline transportation, and X-ray inspection is used to visually detect defects. While manual inspection has traditionally been employed for X-ray defect detection, it suffers from inefficiency and subjectivity. As computer vision techniques have advanced, methods leveraging deep learning have emerged for weld defect detection. However, the practical application of deep learning methods is constrained by their size, limiting their effectiveness in industrial detection tasks. To address these challenges, this paper proposed a lightweight pixel-level automatic detection network for weld defects based on an encoder-decoder structure. Firstly, a lightweight global feature extraction block utilizing self-attention was introduced to enhance detection performance by extracting more discriminative features within defect areas. Subsequently, a high-low level feature fusion block was incorporated to mitigate feature loss resulting from downsampling. Additionally, the depthwise separable convolution was employed to further reduce network size. Our method demonstrated superior detection performance compared to other advanced methods, as evidenced by qualitative comparison and quantitative analysis, with dice coefficient (DICE) and intersection over union (IoU) scores of 0.881 and 0.789, respectively. Moreover, our network exhibited significantly reduced size compared to other advanced detection methods. Thus, the proposed detection network holds considerable promise for effectively detecting weld defects.

A Light-Weight Object Detection Method Based on Knowledge Distillation and Model Pruning for Seam Tracking System

Automatic Seam Detection and Tracking System for Robots Based on Laser Vision

Robotic Seam Tracking System Combining Light-Weight Segmentation Network Design and ADMM-Based Structured Pruning

Robust seam tracking via a deep learning framework combining tracking and detection

Enhanced automatic weld seam tracking: Bayesian optimization and information-based pruning of compact denoising and object detection models

Robotic Seam Tracking System Combining Lightweight Segmentation Network Design and ADMM-Based Structured Pruning

A Robust Detector for Automated Welding Seam Tracking System

Model Optimization and Acceleration Method Based on Meta-Learning and Model Pruning for Laser Vision Weld Tracking System

Light-weight segmentation network based on SOLOv2 for weld seam feature extraction

A Deep Meta-Metric Learning Method for Few-Shot Weld Seam Visual Detection

A Vision Algorithm for Robot Seam Tracking Based on Laser Ranging

Weld Seam Detection Method with Rotational Region Proposal Network.

Research on a Feature Point Detection Algorithm for Weld Images Based on Deep Learning

Robotic Seam Tracking System Combining Convolution Filter and Deep Reinforcement Learning

A weld feature points detection method based on improved YOLO for welding robots in strong noise environment

Robust Welding Seam Tracking and Recognition

A Seam Tracking System Based on a Laser Vision Sensor

A Lightweight Model of Automatic Pixel-Level Detection for Weld Defects

Robust weld seam tracking method based on detection and tracking of laser stripe

A High-Accuracy and Robust Seam Tracking System Based on Adversarial Learning.

A New Welding Seam Recognition Methodology Based on Deep Learning Model MRCNN