Anomaly triplet-net: progress recognition model using deep metric learning considering occlusion for manual assembly work DOI

Takumi Kitsukawa,

Katsutoshi Miura,

Shigeki Yumoto

и другие.

Advanced Robotics, Год журнала: 2024, Номер 39(2), С. 89 - 101

Опубликована: Ноя. 11, 2024

In this paper, we propose a method for recognizing progress in product assembly considering occlusion using deep metric learning. Visualizing the process factories is crucial enhancing work efficiency and minimizing disposal costs. However, there problem that products are managed by pasting them on paper with status written them. We solve of having to manage manually. First, target detected from images acquired fixed-point camera installed factory learning-based object detection method. Next, area cropped image. Finally, classification based learning image, estimated as rough step. As specific estimation model, an Anomaly Triplet-Net which improved model existing Triplet Loss. This considers anomaly samples. experiments, 82.9 [%] success rate achieved Triplet-Net. also experimented practicality sequence detection, cropping, progression estimation, confirmed effectiveness overall system.

Язык: Английский

A DSF-net-based approach to dual-branch instance segmentation of weak bridge defects DOI
He Zhang,

Ruihong Shen,

Jiawei Lei

и другие.

Engineering Structures, Год журнала: 2025, Номер 327, С. 119583 - 119583

Опубликована: Янв. 4, 2025

Язык: Английский

Процитировано

0

Application of machine learning to leakage detection of fluid pipelines in recent years: A review and prospect DOI

Jianwu Chen,

Xiao Wu, Zhibo Jiang

и другие.

Measurement, Год журнала: 2025, Номер unknown, С. 116857 - 116857

Опубликована: Янв. 1, 2025

Язык: Английский

Процитировано

0

Inner wall defect detection in oil and gas pipelines using point cloud data segmentation DOI
Zhouyu Yan, Hong Zhao

Automation in Construction, Год журнала: 2025, Номер 173, С. 106098 - 106098

Опубликована: Март 4, 2025

Язык: Английский

Процитировано

0

Corrosion Depth Prediction of a Buried Pipeline Based on TEM and MSDBO-BiLSTM DOI

Fulin Song,

Hong Zhao, Xingyuan Miao

и другие.

Journal of Pipeline Systems Engineering and Practice, Год журнала: 2025, Номер 16(2)

Опубликована: Март 12, 2025

Язык: Английский

Процитировано

0

Research of Highway Bridge Settlement Monitoring Technology based on Machine Vision DOI Creative Commons
Zhao Qian,

Chunhao Hu,

Guoqing Xia

и другие.

Journal of Research in Science and Engineering, Год журнала: 2024, Номер 6(7), С. 29 - 32

Опубликована: Июль 28, 2024

In view of the significant impact deep foundation pit excavation on surface surrounding roads and bridges, widely used monitoring technology still relies manual detection means, which leads to consumption a large number human material resources, efficiency is relatively low. Therefore, this paper provides method system highway bridge pile displacement based machine vision. Through real-time automatic settlement changes, it targeted suggestions guidance for maintenance during excavation. At same time, new type marker module provided enhance accuracy feature point recognition in image processing. The results show that vision can automatically monitor real time with high accuracy, improve safety stability bridges construction.

Язык: Английский

Процитировано

1

A New Hybrid Velocity Prediction Model for Pipeline Detectors Based on Bo-Ssa, Bilstm, and the Attention Mechanism DOI
Junjie Ma, Yiming Li, Zhongchao Zhang

и другие.

Опубликована: Янв. 1, 2024

Язык: Английский

Процитировано

0

Identification of Coating Layer Pipeline Defects Based on the GA-SENet-ResNet18 Model DOI

S.. Wang,

Wei Liang,

Fang Shi

и другие.

International Journal of Pressure Vessels and Piping, Год журнала: 2024, Номер unknown, С. 105327 - 105327

Опубликована: Сен. 1, 2024

Язык: Английский

Процитировано

0

Anomaly triplet-net: progress recognition model using deep metric learning considering occlusion for manual assembly work DOI

Takumi Kitsukawa,

Katsutoshi Miura,

Shigeki Yumoto

и другие.

Advanced Robotics, Год журнала: 2024, Номер 39(2), С. 89 - 101

Опубликована: Ноя. 11, 2024

In this paper, we propose a method for recognizing progress in product assembly considering occlusion using deep metric learning. Visualizing the process factories is crucial enhancing work efficiency and minimizing disposal costs. However, there problem that products are managed by pasting them on paper with status written them. We solve of having to manage manually. First, target detected from images acquired fixed-point camera installed factory learning-based object detection method. Next, area cropped image. Finally, classification based learning image, estimated as rough step. As specific estimation model, an Anomaly Triplet-Net which improved model existing Triplet Loss. This considers anomaly samples. experiments, 82.9 [%] success rate achieved Triplet-Net. also experimented practicality sequence detection, cropping, progression estimation, confirmed effectiveness overall system.

Язык: Английский

Процитировано

0