Detection method of small size defects on pipeline weld surface based on improved YOLOv7
Open Access
- 12 December 2024
- journal article
- physical sciences
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (12) , e0313348
Abstract
The background of pipeline weld surface defect image is complex, and the defect size is small. Aiming at the small defect size in the weld image, which is easy to cause missed detection and false detection, a lightweight target detection algorithm based on improved YOLOv7 is proposed. Firstly, in the feature fusion network of YOLOv7, the detection ability of the algorithm to detect small and medium-sized targets in defect images is enhanced by adding a 160*160 small target detection head. Then, the convolution module in the backbone network and the feature fusion network is replaced by the depthwise separable convolution with less computational overhead, so as to effectively reduce the network calculation, parameter quantity and model volume. Finally, the loss function CIoU of YOLOv7 is optimized to EIoU loss function to accelerate the convergence speed of the model. The experimental results show that the defect detection mAP@0.5 based on the improved YOLOv7 algorithm can reach 72.2%, which is 11% higher than that of YOLOv7, and the model calculation amount and parameter amount are reduced by 75.6% and 60.3%, respectively. It can completely detect the small size defects and has a high degree of confidence, which can be effectively applied to the detection of small size defects on the surface of pipeline weld.Keywords
This publication has 20 references indexed in Scilit:
- YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors2023
- LFF-YOLO: A YOLO Algorithm With Lightweight Feature Fusion Network for Multi-Scale Defect DetectionIEEE Access, 2022
- Multiple attentional path aggregation network for marine object detectionApplied Intelligence, 2022
- An improved feature pyramid network for object detectionNeurocomputing, 2022
- Development of a YOLO-V3-based model for detecting defects on steel strip surfaceMeasurement, 2021
- DF-SSD: An Improved SSD Object Detection Algorithm Based on DenseNet and Feature FusionIEEE Access, 2020
- A deep-learning-based approach for fast and robust steel surface defects classificationOptics and Lasers in Engineering, 2019
- Facial Expression Recognition with Faster R-CNNProcedia Computer Science, 2017
- SSD: Single Shot MultiBox Detector2016
- Reliability of non-destructive test techniques in the inspection of pipelines used in the oil industryInternational Journal of Pressure Vessels and Piping, 2008