Figure 5 - uploaded by Zhang Jiaqiao
Content may be subject to copyright.
Source publication
ABSTRACT
The steel strip is an essential raw material in the machinery industry. Besides, the surface defects of the steel strip directly determine its performance. To achieve rapid and effective detection of the defects, a
CP-YOLOv3-dense (classification priority YOLOv3 DenseNet) neural
network was proposed in the present study. The model used YO...
Contexts in source publication
Context 1
... entire dataset contains over 5000 ground truth boxes. Figure 5 presents the examples of annotated defect images in the NEU-DET dataset. ...
Similar publications
The traditional switchgear partial discharge pattern recognition method lacks a certain generalization performance and the recognition accuracy is low, which is difficult to apply in practical engineering. In view of this, this paper proposes a switchgear partial discharge pattern recognition method based on residual convolutional neural network. B...
Citations
... Moreover, experimental results demonstrate that this network can achieve a reasonable detection rate in complex environments. To enhance the performance of surface defect detection on steel strips, a classification priority YOLOv3-dense network is proposed [25]. ...
Efficient detection of surface defects is primary for ensuring product quality during manufacturing processes. To enhance the performance of deep learning‐based methods in practical applications, the authors propose Dense‐YOLO, a fast surface defect detection network that combines the strengths of DenseNet and you only look once version 3 (YOLOv3). The authors design a lightweight backbone network with improved densely connected blocks, optimising the utilisation of shallow features while maintaining high detection speeds. Additionally, the authors refine the feature pyramid network of YOLOv3 to increase the recall of tiny defects and overall positioning accuracy. Furthermore, an online multi‐angle template matching technique is introduced based on normalised cross‐correlation to precisely locate the detection area. This refined template matching method not only accelerates detection speed but also mitigates the influence of the background. To validate the effectiveness of our enhancements, the authors conduct comparative experiments across two private datasets and one public dataset. Results show that Dense‐YOLO outperforms existing methods, such as faster R‐CNN, YOLOv3, YOLOv5s, YOLOv7, and SSD, in terms of mean average precision (mAP) and detection speed. Moreover, Dense‐YOLO outperforms networks inherited from VGG and ResNet, including improved faster R‐CNN, FCOS, M2Det‐320 and FRCN, in mAP.
... Cheng et al. [20] enhanced RetinaNet by proposing a network with differential channel attention and adaptive spatial feature fusion, which can better merge features at different scales and improve the accuracy of defect detection. Zhang et al. [21] proposed YOLOv3-DenseNet using image priority classification, which can obtain multilevel convolutional features from the output of densely connected blocks to improve feature fusion and thus further improve detection accuracy. Ge et al. [22] proposed YOLOX, which incorporates the concept of anchor-free [23] to greatly improve the prediction accuracy of the model for objects. ...
Steel is a fundamental material in the manufacturing process, and the quality of the steel used directly affects the quality of the final product. During the manufacturing process, a variety of complex and irregular defects may form on the surface of the steel. In order to detect these defects, this paper proposes the DF-YOLOv7 model. The model employs the K-means + + algorithm to adjust the anchor box sizes across datasets, thereby enhancing the extraction of features for different defects. Furthermore, the D-SPPCSPC module is employed to enhance defect detection and reduce model parameters. Additionally, the CIoU Loss with Focal module addresses positive–negative sample imbalance by focusing on high-quality anchor boxes. Experimental results demonstrate that the proposed model achieves an mAP of 0.771 on the NEU-DET dataset, representing a 3.6% improvement over the original model. It outperforms some state-of-the-art detectors and meets the real-time industrial detection requirements.
... The results show that the model achieves 0.805 mAP and shows high efficiency. Zhang et al. [131] proposed a CP-YOLOv3-dense neural network for fast and effective detection of rigid band defects. The model used YOLOv3 as the basic network to prioritize the image, and then replaced the two residual network modules with two dense network modules. ...
In recent years, the development of machine vision research is rapid in several areas. In order to promote the better development of machine vision research, it is necessary to clarify its development and application direction. At present, there are few reviews on the application direction of machine vision. This paper sorts out the application of machine vision in various fields, and summarizes the current application status of machine vision from four main functions: recognition, measurement, classification and detection. This paper mainly introduces the improvement of different algorithms of machine vision and its application in medical, agriculture, manufacturing and other industries, providing guidance for the selection of machine vision research direction.
... This makes it possible to quickly identify flaws and take swift corrective action to reduce production losses. In industrial contexts, image processing techniques are essential for automating the detection of steel faults and enhancing the speed, accuracy, and efficiency of inspections [9]. Image processing makes early flaw detection and mitigation possible by utilizing the power of computational analysis, resulting in steel products of greater quality. ...
Accurate detection of defects on steel surfaces is crucial for maintaining quality standards in steel production. This paper addresses the challenge of classifying steel sheets into distinct defect categories by presenting a robust method that leverages deep learning and advanced optimization techniques. We propose a novel approach that utilizes the ResNet101 model to extract deep features, which are then classified using a support vector machine (SVM). To enhance the SVM's performance, Bayesian optimization is employed for hyperparameter tuning. Our method is validated using the "Severstal: Steel Defect Detection" dataset from Kaggle, achieving a validation accuracy of 89.1% and a test accuracy of 90.6%, with a classification error of 0.10934. Additionally, the area under the curve (AUC) for each class exceeds 0.95 in both the validation and test sets, demonstrating excellent discriminatory power. Further evaluation on the DAGM dataset achieved flawless results, with an accuracy of 100%, AUC of 1, sensitivity of 100%, specificity of 100%, precision of 100%, MCC of 100%, F1 score of 1, and kappa of 100%. On the NEU dataset, our method achieved an accuracy of 97.92%, sensitivity of 97.92%, specificity of 99.58%, precision of 98.06%, F1 score of 0.9791, MCC of 97.55%, and kappa of 92.50%. These results demonstrate the robustness and adaptability of the proposed method, offering an efficient and reliable solution for automating steel defect detection and surface defect classification in industrial applications.
... In 2020, Zhang et al. (2020) developed Classification Priority YOLOv3 DenseNet (CP-YOLOv3-Dense), a variant designed for detecting surface defects on steel, building on the YOLOv3 architecture. They replaced two residual network modules with Densely Connected Convolutional Network (DenseNet) to enhance feature fusion utilization. ...
One of the focal points in industrial product defect detection lies in the utilization of deep learning-based object detection algorithms. With the continuous introduction of these algorithms and their refined models, notable achievements have been attained. However, challenges persist in industrial settings, such as substantial variations in defect scales, the delicate balance between accuracy and speed, and the detection of small objects. Various methods have been proposed to address these challenges and propel the advancement of defect detection. To comprehensively review the latest developments in deep learning-based industrial product defect detection algorithms and foster further progress, this paper encompasses typical datasets and evaluation metrics used in industrial product defect detection, traces the development history of supervised one-stage and two-stage object detection algorithm-based and unsupervised algorithm-based industrial defect detection methods, discusses major challenges, and outlines future directions. It highlights the potential for further improving the accuracy, speed, and reliability of defect detection systems in industrial applications.
... Next, Sun et al. proposed a detection algorithm based on the adaptive threshold Canny algorithm, which no longer required manual threshold setting [6]. But when the light or noise in the image is not uniform, it still cannot produce a stable effect [7][8][9][10][11][12][13]. Subirats et al. proposed a crack detection method based on continuous wavelet transform, which could effectively detect and segment crack images but still had some noise [14]. ...
Timely discovery and disposal of road risk sources constitute the cornerstone of road operation safety. Presently, the detection of road risk sources frequently relies on manual inspections via inspection vehicles, a process that is both inefficient and time-consuming. To tackle this challenge, this paper introduces a novel automated approach for detecting road risk sources, termed the multi-scale lightweight network (MSLN). This method primarily focuses on identifying road surfaces, potholes, and scattered objects. To mitigate the influence of real-world factors such as noise and uneven brightness on test results, pavement images were carefully collected. Initially, the collected images underwent grayscale processing. Subsequently, the median filtering algorithm was employed to filter out noise interference. Furthermore, adaptive histogram equalization techniques were utilized to enhance the visibility of cracks and the road background. Following these preprocessing steps, the MSLN model was deployed for the detection of road risk sources. Addressing the challenges associated with two-stage network models, such as prolonged training and testing times, as well as deployment difficulties, this study adopted the lightweight feature extraction network MobileNetV2. Additionally, transfer learning was incorporated to elevate the model’s training efficiency. Moreover, this paper established a mapping relationship model that transitions from the world coordinate system to the pixel coordinate system. This model enables the calculation of risk source dimensions based on detection outcomes. Experimental results reveal that the MSLN model exhibits a notably faster convergence rate. This enhanced convergence not only boosts training speed but also elevates the precision of risk source detection. Furthermore, the proposed mapping relationship coordinate transformation model proves highly effective in determining the scale of risk sources.
... As one of the common metals in the manufacturing industry, steel has been applied in various fields such as aerospace, machinery manufacturing, and petrochemicals with the development of national industrialisation [1][2][3]. However, in the process of steel production and manufacturing, various factors such as production equipment and adverse environments can cause defects such as cracks and scratches on the surface of the steel [4][5][6], leading to a decrease in its corrosion resistance and affecting its quality [7][8][9]. ...
... The object detection accuracy was achieved on PASCAL VOC 2007 (73.2% mAP) and 2012 (70.4% mAP). Zhang et al. [11] proposed the classification priority YOLOv3 densenet neural network to detect the surface defects of the steel strips. The improved network successfully identified six types of surface defects with an impressive detection precision of 85.7%. ...
Micro-electro-mechanical system (MEMS) sensors have been widely used in the automotive field owing to their small size, low cost, high reliability. Various defects caused by complicated environment and multi-step process in the manufacturing process of automotive MEMS pressure sensor. These defects are often small in size but can significantly impact the performance of products. The traditional detection method is inefficient and inaccurate. In this study, an improved you only look once v8 (YOLOv8) CR network is proposed to detect the defects of the automotive MEMS pressure sensors. The network is an improved version of YOLOv8. The faster version of CSP bottleneck with two convolutions (C2f) and receptive-field convolutional block attention (RFCBAMConv) module are applied. The ablation experiments confirm the impact of C2f and RFCBAMConv modules on network performance. The improved YOLOv8-CR network can effectively enhance the accuracy and speed of defect detection of the automotive MEMS pressure sensor. Furthermore, comparisons are made between the improved YOLOv8-CR network and other networks such as YOLOv8-large, YOLOv5, and Fast-RCNN, demonstrating its effectiveness in identifying five types of defects in the automotive MEMS pressure sensors. The improved YOLOv8-CR exhibits considerably superior defect detection capabilities with a mean average precision (mAP) of 94.98% and a single picture detection time of 42.68 ms on the test set. The improved YOLOv8-CR network is helpful to realize real-time and accurate online monitoring of product quality in the key process of the automotive MEMS pressure sensors.
... Classification is a supervised machine learning approach used to solve various problems [2,10]. In recent times, with advancements in Computer Vision (CV) [11] and Artificial Intelligence (AI) technologies, various types of Statistical [12,13], Machine Learning (ML) [13,14], and Deep Learning (DL) [2,4,5,15] methods have been developed for feature extraction and automatic classification of steel surface defects. However, despite these developments, none of these methods can achieve outstanding performance across all problems and scenarios due to insufficient and imbalanced data. ...
Smart manufacturing is a process that optimizes factory performance and production quality by utilizing various technologies including the Internet of Things (IoT) and artificial intelligence (AI). Quality control is an important part of today’s smart manufacturing process, effectively reducing costs and enhancing operational efficiency. As technology in the industry becomes more advanced, identifying and classifying defects has become an essential element in ensuring the quality of products during the manufacturing process. In this study, we introduce a CNN model for classifying defects on hot-rolled steel strip surfaces using hybrid deep learning techniques, incorporating a global average pooling (GAP) layer and a machine learning-based SVM classifier, with the aim of enhancing accuracy. Initially, features are extracted by the VGG19 convolutional block. Then, after processing through the GAP layer, the extracted features are fed to the SVM classifier for classification. For this purpose, we collected images from publicly available datasets, including the Xsteel surface defect dataset (XSDD) and the NEU surface defect (NEU-CLS) datasets, and we employed offline data augmentation techniques to balance and increase the size of the datasets. The outcome of experiments shows that the proposed methodology achieves the highest metrics score, with 99.79% accuracy, 99.80% precision, 99.79% recall, and a 99.79% F1-score for the NEU-CLS dataset. Similarly, it achieves 99.64% accuracy, 99.65% precision, 99.63% recall, and a 99.64% F1-score for the XSDD dataset. A comparison of the proposed methodology to the most recent study showed that it achieved superior results as compared to the other studies.
... Kou et al. [16] effectively enhances the extraction of feature information on the surface of steel strips by adding dense convolutional blocks into YOLOv3 [17]. Zhang et al. [18] replaces two residual network modules with two dense network modules, which further enhances the feature extraction. However, due to the complex background and variable location, multi-scale feature extraction may still miss critical information in the detection of surface defects on steel strips. ...
The surface quality of steel strip is a critical indicator of the quality of hot‐rolled strip, so accurate inspection of its surface is essential. However, the complex texture of steel strip surface defects makes the detection challenging. Here, a multi‐scale feature fusion and attention based YOLOv5 (MFFA‐YOLOv5) model is proposed. Specifically, the bottom layer features are up‐sampled and fused not only with the middle layer features, but also with the top layer features, so that the model better captures the surface texture information of the steel strip. Secondly, an improved attention mechanism module is introduced to deal with the global and local information of the steel strip surface by introducing down‐sampling and up‐sampling paths based on Convolutional Block Attention Module (CBAM). Meanwhile, a self‐attention mechanism path is added to improve the capability of feature representation. Experimental results on the NEU‐DET dataset show that the MFFA‐YOLOv5 model significantly outperforms other state‐of‐the‐art methods.
