Computer Vision Techniques in Manufacturing

Zhou, Longfei; Zhang, Linlin; Konz, Nicholas

doi:10.1109/tsmc.2022.3166397

Cited by 77 publications

(17 citation statements)

References 151 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The acquired visual data and data labeling information are effectively analyzed, processed and learned, which is the need for machine learning, deep learning and other methods to construct appropriate target detection models. However, target detection in cross-modal scenes with low contrast, cluttered backgrounds, cloud interference and large target scale variations in the context of visual big data puts higher requirements on algorithm performance [3][4][5]. In order to be able to improve the efficiency and performance of multi-source visual information processing more effectively, many scholars have proposed target detection algorithms based on convolutional neural network (CNN) [6][7][8].…”

Section: Index Terms-target Tracking; Transformer; Efficient Self-att...mentioning

confidence: 99%

Target Detection Algorithm Based on Efficient Self-Attention-Convolution Enhanced Transformer

An,

Wang,

Liu

2024

Preprint

View full text Add to dashboard Cite

Since the target detection algorithm based on convolutional neural network suffers from limited convolutional kernel receptive field, it leads to the model's inability to perceive the remote semantic information in the image. Because the Transformer model does not have the limitation of local receptive fields, it is introduced into the field of target detection, and many scholars have proposed target detection algorithms based on Transformer and its variants. However, the Transformer model has the difficulties of not being able to extract the deep feature information and the high computational complexity of the standard self-attention mechanism in the process of target detection and recognition applications. Aiming at the above two core problems, we have carried out in-depth analysis and research, and proposed an encoder-decoder model consisting of a convolutional layer and a Transformer module. And then, we constructed the efficient multi-head self-attention mechanism, which can capture both local and remote contextual information of target image features. Then, we design efficient convolutional module-enhanced cross-window connectivity, which can significantly improve the characterization and global modeling capabilities of Transformer model. In addition, we propose the convolution-enhanced Transformer learning framework, which improves the adaptability to different datasets, which also integrates the sparse sampling strategy. It can significantly reduce the memory and computational requirements in large-scale image processing. Finally, we propose a target detection algorithm based on a new Transformer framework. We conducted ablation experiments and computational performance comparison experiments on several HRRS scenes and natural scene datasets. The experimental results confirm that our proposed method obtains optimal results in terms of weighted F-measure, average F-measure and MAE. Moreover, our proposed method has clearer edge information and more accurate target localization information in the visual effect of detection results.

show abstract

Section: Index Terms-target Tracking; Transformer; Efficient Self-att...mentioning

confidence: 99%

Target Detection Algorithm Based on Efficient Self-Attention-Convolution Enhanced Transformer

An,

Wang,

Liu

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…A computer vision model is developed to automatically count the number of workpieces from a video. The framework is based on a Manufacturing-oriented CV system by Zhou et al [18] as illustrated in Fig. 1.…”

Section: Framework Of the Proposed Counting Systemmentioning

confidence: 99%

A Computer Vision-Based System for Metal Sheet Pick Counting

Ji¹,

Pannakkong²,

Buddhakulsomsiri³

2023

Computers, Materials &Amp; Continua

View full text Add to dashboard Cite

Inventory counting is crucial to manufacturing industries in terms of inventory management, production, and procurement planning. Many companies currently require workers to manually count and track the status of materials, which are repetitive and non-value-added activities but incur significant costs to the companies as well as mental fatigue to the employees. This research aims to develop a computer vision system that can automate the material counting activity without applying any marker on the material. The type of material of interest is metal sheet, whose shape is simple, a large rectangular shape, yet difficult to detect. The use of computer vision technology can reduce the costs incurred from the loss of high-value materials, eliminate repetitive work requirements for skilled labor, and reduce human error. A computer vision system is proposed and tested on a metal sheet picking process for multiple metal sheet stacks in the storage area by using one video camera. Our results show that the proposed computer vision system can count the metal sheet picks under a real situation with a precision of 97.83% and a recall of 100%.

show abstract

“…The utilization of Artificial Intelligence-based Computer Vision models (AI-CV model) can support a wide range of processes and applications along the life cycle of technical systems such as machines and plants by detecting components in photo and video data (Zhou et al, 2023). Current research on AI-CV models in engineering focuses on the detection of object categories like screws, bearings or pipes by utilizing training data sets with generic categories (Drost et al, 2017).…”

Section: Introductionmentioning

confidence: 99%

Towards a process for the creation of synthetic training data for AI-computer vision models utilizing engineering data

Schwoch,

Dammann,

Bartl

et al. 2024

Proc. Des. Soc.

View full text Add to dashboard Cite

Artificial Intelligence-based Computer Vision models (AI-CV models) for object detection can support various applications over the entire lifecycle of machines and plants such as monitoring or maintenance tasks. Despite ongoing research on using engineering data to synthesize training data for AI-CV model development, there is a lack of process guidelines for the creation of such data. This paper proposes a synthetic training data creation process tailored to the particularities of an engineering context addressing challenges such as the domain gap and methods like domain randomization.

show abstract

Computer Vision Techniques in Manufacturing

Abstract: This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Cited by 77 publications

References 151 publications

Target Detection Algorithm Based on Efficient Self-Attention-Convolution Enhanced Transformer

Target Detection Algorithm Based on Efficient Self-Attention-Convolution Enhanced Transformer

A Computer Vision-Based System for Metal Sheet Pick Counting

Towards a process for the creation of synthetic training data for AI-computer vision models utilizing engineering data

Contact Info

Product

Resources

About