Enhanced Lightweight Network for Melt Pool Contour Extraction with ECA Attention Mechanism
- Vol. 54, Issue 12, Pages: 28-34(2024)
Published: 25 December 2024
DOI: 10.7512/j.issn.1001-2303.2024.12.05
移动端阅览
浏览全部资源
扫码关注微信
Published: 25 December 2024 ,
移动端阅览
张昆,范冬阳,袁飞,等.ECA注意力机制增强的轻量级网络焊接熔池轮廓提取方法[J].电焊机,2024,54(12):28-34.
ZHANG Kun, FAN Dongyang, YUAN Fei, et al.Enhanced Lightweight Network for Melt Pool Contour Extraction with ECA Attention Mechanism[J].Electric Welding Machine, 2024, 54(12): 28-34.
熔池形貌特征在线检测是实现机器人焊接智能化与自动化的重要途径之一。然而,目前的熔池轮廓提取方法多使用基于阈值分割的传统处理算法,这类算法在弧光干扰的情况下,熔池轮廓准确度低、鲁棒性差;而基于深度学习对熔池轮廓进行提取的报道还较少,并且在实时性方面存在一定的挑战。针对这一问题,本文提出一种改进的DeepLabV3 +的高效语义分割模型。该模型采用轻量级的 MobileNetv2 主干网络,降低模型复杂度,提高推理速度;并在 ASPP 特征提取模块后添加 ECA 注意力机制,增强网络对熔池特征的关注,提升分割精度;针对熔池图像数据集中前景和背景像素数量不均衡的问题,采用 Dice Loss 和交叉熵损失函数的线性组合作为新的损失函数,改善模型训练效果。实验结果表明,该模型在熔池轮廓提取任务中性能优异,平均交并比(mIoU)达到96.08% ,类别平均像素准确率(mPA)为97.85%,推理时间由60.72 ms缩短到23.11 ms,满足了熔池轮廓提取准确性与实时性的要求。
Online detection of arc additive molten pool morphology features is an important way to achieve system intelligence and automation. However
current methods for extracting molten pool contours mostly use traditional processing algorithms based on threshold segmentation. These algorithms have low accuracy and poor robustness in the presence of arc interference. Reports on using deep learning for molten pool contour extraction are relatively rare
and there are certain challenges in terms of real-time performance. To address this issue
this paper proposes an improved efficient semantic segmentation model based on DeepLabv3+ for molten pool contour extraction. Experimental results show that the proposed algorithm achieves the mean Intersection over Union (mIoU) of 96.08% and the mean pixel accuracy (mPA) of 97.85%
while reducing inference time from 60.72 ms to 23.11 ms
thus meeting the requirements for accurate and real-time molten pool contour extraction.
语义分割熔池轮廓提取轻量级网格ECA注意力机制
semantic segmentationmelt pool contour extraction methodlightweight network
杨启,田虎成,闫昭华. 激光近净成形中熔池宽度实时监控系统的研究[J]. 激光与红外,2019,49(09):1060-1067.
YANG Q,TIAN H C,YAN S H. Study on real-time monitoring system of molten pool width during laser engineered net shaping[J]. LASER & INFRARED,2019,49(09):1060-1067.
洪宇翔,杨明轩,都东. 铝合金爬坡TIG焊熔池失稳状态的视觉检测[J]. 焊接学报,2021,42(10):8-13.
HONG Y X,YANG M X,DU D. Unstable state vision detection of molten pool during aluminum alloy climbing-TIG welding[J]. Transactions of the China Welding Institution,2021,42(10):8-13.
Fang J,Wang K. Weld Pool Image Segmentation of Hump Formation Based on Fuzzy C-Means and Chan-Vese Model[J]. Journal of Materials Engineering and Performance,2019,28(7):4467-4476.
张楚昊,赵壮,陆骏. 基于边缘导向算子模板匹配的熔池轮廓提取方法[J]. 焊接学报,2022,43(02):67-74+118.
ZHANG C H,ZHAO Z,LU J. Molten pool contour extraction method based on edge oriented operator template matching[J]. Transactions of the China Welding Institution,2022,43(2):67-74.
Liu W,Wang Z,Tian L,et al. Melt pool segmentation for additive manufacturing:A generative adversarial network approach[J]. Computers & Electrical Engineering,2021,92:107183.
Cai W,Jiang P,Shu L,et al. Real-time identification of molten pool and keyhole using a deep learning-based semantic segmentation approach in penetration status monitoring[J]. Journal of manufacturing processes,2022(4):76.
Chen L C,Zhu Y K,Papandreou G,et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]//Computer Vision-ECCV,2018:833-851.
Sandler M,Howard A,Zhu M,et al. MobileNetv2: Inverted residuals and linear bottlenecks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR),2018:4510-4520.
Lennart B,Aurelio C.Generalized attention-weighted reinforcement learning[J]. Neural Networks,2022,145:10-21.
Wang Q,Wu B,Zhu P,et al. ECA: Efficient Channel Attention for Deep Convolutional Neural Networks[C]//IEEE Conference on Computer Vision and Pattern Recognition,2020.
Goh T Y,Basah S N,Yazid H,et al. Performance analysis of image thresholding: Otsu technique[J]. Measurement,2018,114:298-307.
Kaur P,Maini R. Performance Evaluation of various thresholding methods using canny edge detector[J]. International Journal of Computer Applications,2013,71(9):26-32.
Weng Y,Zhou T,Li Y,et al. Nas-unet: Neural architecture search for medical image segmentation[J]. IEEE access,2019,7:44247-44257.
Yang C,Guo H. A Method of Image Semantic Segmentation Based on PSPNet[J]. Mathematical Problems in Engineering,2022(1):8958154.
相关文章
相关作者
相关机构