Chunmei Duan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Chunmei Duan has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Taochuan Zhang
College of Mechanical and Electrical Engineering, Foshan Polytechnic, Foshan, China
3 shared publications - Huilin Tan
College of Information Science and Engineering, Shaoyang University, Shaoyang, China
2 shared publications - Gong Wang
Center for Advanced Laser Technology, Hebei University of Technology, Tianjin, China
1 shared publication Haibo Zhou
College of Mechanical and Electrical Engineering, Central South University, Changsha, China
1 shared publication- Lei Han
School of Automobile and Transportation Engineering, Guangdong Polytechnic Normal University, Guangzhou, 510665, China
1 shared publication Liang Zhang
College of Mechanical and Electrical Engineering, Central South University, Changsha, China
1 shared publication
Latest Publications
- Open Access
- 16 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 134312-134324
Research on Surface Defects Classification for PET Preform by Fusing Multi-Scale Features
Product quality control is an important component of PET preform production process. In order to improve the accuracy and reliability of PET preform detection, reduce the cost of product quality defects, and improve the overall level of product quality, this paper proposes a surface defect detection method for PET preform based on multi-scale feature fusion. Multi-scale features fusion combines features from different scales to produce more accurate and robust feature representations, which improve the accuracy, stability and adaptability of PET preform detection model. This feature processing model constructs a three-branch network, generating feature maps of different sizes. Then, the low-level feature, the middle-level feature and the high-level feature are fused to obtain a PET preform feature map based on multi-scale feature fusion by using a maximum value fusion strategy. The original image and feature map of the PET preform are respectively input into the deep convolutional neural network for feature extraction and fusion. And finally the fusion features are input into the ECOC-SVM to achieve surface defect classification of the PET preform. The machine vision platform is designed for collecting and detecting PET preform data. By importing the model designed in this paper, surface defect recognition and defect category judgment of PET preform can be achieved. Through experimental data analysis, the higher testing detection accuracy of this method is 98.8%, with an average testing accuracy of 97.1%.
- 16 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 134312-134324
Open AccessResearch on Surface Defects Classification for PET Preform by Fusing Multi-Scale Features
Product quality control is an important component of PET preform production process. In order to improve the accuracy and reliability of PET preform detection, reduce the cost of product quality defects, and improve the overall level of product quality, this paper proposes a surface defect detection method for PET preform based on multi-scale feature fusion. Multi-scale features fusion combines features from different scales to produce more accurate and robust feature representations, which improve the accuracy, stability and adaptability of PET preform detection model. This feature processing model constructs a three-branch network, generating feature maps of different sizes. Then, the low-level feature, the middle-level feature and the high-level feature are fused to obtain a PET preform feature map based on multi-scale feature fusion by using a maximum value fusion strategy. The original image and feature map of the PET preform are respectively input into the deep convolutional neural network for feature extraction and fusion. And finally the fusion features are input into the ECOC-SVM to achieve surface defect classification of the PET preform. The machine vision platform is designed for collecting and detecting PET preform data. By importing the model designed in this paper, surface defect recognition and defect category judgment of PET preform can be achieved. Through experimental data analysis, the higher testing detection accuracy of this method is 98.8%, with an average testing accuracy of 97.1%.
- Open Access
- 26 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 61512-61528
RAFBSD: An Efficient Detector for Accurate Identification of Defects in Photovoltaic Cells
Currently, defect detection for photovoltaic (PV) electroluminescence (EL) images faces three challenges: limited training data and complex backgrounds result in low accuracy in detecting defects; the diverse shapes of specific defects often lead to frequent false alarms; and existing models still require improvement in accurately recognizing these 12 specific defects. An innovative detector, called the Reconfigurable Adaptive Focus Background Suppression Detector (RAFBSD), is proposed to tackle these challenges. This approach utilizes the advanced Deformable Attention Transformer (DAT) technology to effectively focus on regions with defects. By doing so, it captures more informative features while significantly reducing noise interference. Additionally, the Omni-dimensional Dynamic Convolution (ODConv) and the Receptive-Field Attention convolutional (RFAConv) modules enhance accuracy by effectively allocating attention weights to optimize model performance. Compared to YOLOv8, RAFBSD has demonstrated substantial enhancements in both mAP50 (by 4.7%) and mAP50-95 (by 5.3%). Furthermore, RAFBSD demonstrates an outstanding capability to precisely identify flaws in 12 different classifications of PVEL images, exceeding the recognition capacity of YOLOv8 models by three additional types. These findings unequivocally establish RAFBSD as an efficient and groundbreaking object detection model.
- 26 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 61512-61528
Open AccessRAFBSD: An Efficient Detector for Accurate Identification of Defects in Photovoltaic Cells
Currently, defect detection for photovoltaic (PV) electroluminescence (EL) images faces three challenges: limited training data and complex backgrounds result in low accuracy in detecting defects; the diverse shapes of specific defects often lead to frequent false alarms; and existing models still require improvement in accurately recognizing these 12 specific defects. An innovative detector, called the Reconfigurable Adaptive Focus Background Suppression Detector (RAFBSD), is proposed to tackle these challenges. This approach utilizes the advanced Deformable Attention Transformer (DAT) technology to effectively focus on regions with defects. By doing so, it captures more informative features while significantly reducing noise interference. Additionally, the Omni-dimensional Dynamic Convolution (ODConv) and the Receptive-Field Attention convolutional (RFAConv) modules enhance accuracy by effectively allocating attention weights to optimize model performance. Compared to YOLOv8, RAFBSD has demonstrated substantial enhancements in both mAP50 (by 4.7%) and mAP50-95 (by 5.3%). Furthermore, RAFBSD demonstrates an outstanding capability to precisely identify flaws in 12 different classifications of PVEL images, exceeding the recognition capacity of YOLOv8 models by three additional types. These findings unequivocally establish RAFBSD as an efficient and groundbreaking object detection model.
- Open Access
- 16 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 9, 156973-156986
Multi-Stream Deep Convolutional Neural Network for PET Preform Surface Defects Detection and Classification
Due to the influence of technology factors, various defects will appear in the production process of PET (Polyethylene Terephthalate) preform, and affect the PET preform quality. In order to meet the requirements of the quality inspection efficiency and accuracy for PET preform, a novel multi-stream deep CNN (Convolutional Neural Network) model is performed to effectively identify and classify PET preform surface defects for the first time in this paper. In recent years, the method based on multi-stream feature fusion has a good application prospect in defect detection. In order to solve the problem of PET preform surface defects detection and classification, we study the deep CNN structure and propose a multi-stream deep network structure, which makes full use of the rich multi-feature information from different network structures. Two different deep CNNs are respectively trained to extract features from the original image, and one of them is trained to extract feature from the corresponding gradient image. We adopt wavelet transform fusion strategy to realize feature fusion and input them into ECOC-SVM (a classified Error Correcting Output Codes model using Support Vector Machines as binary learner) model for detection and classification. Using Bayesian optimization function to optimize the hyperparameters is performed to choose the best performance configuration. Through experimental analyses, the higher detection accuracy of the proposed method is 98.5%, and it is proved that the proposed model has good convergence, accuracy, stability and generalization ability.
- 16 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 9, 156973-156986
Open AccessMulti-Stream Deep Convolutional Neural Network for PET Preform Surface Defects Detection and Classification
Due to the influence of technology factors, various defects will appear in the production process of PET (Polyethylene Terephthalate) preform, and affect the PET preform quality. In order to meet the requirements of the quality inspection efficiency and accuracy for PET preform, a novel multi-stream deep CNN (Convolutional Neural Network) model is performed to effectively identify and classify PET preform surface defects for the first time in this paper. In recent years, the method based on multi-stream feature fusion has a good application prospect in defect detection. In order to solve the problem of PET preform surface defects detection and classification, we study the deep CNN structure and propose a multi-stream deep network structure, which makes full use of the rich multi-feature information from different network structures. Two different deep CNNs are respectively trained to extract features from the original image, and one of them is trained to extract feature from the corresponding gradient image. We adopt wavelet transform fusion strategy to realize feature fusion and input them into ECOC-SVM (a classified Error Correcting Output Codes model using Support Vector Machines as binary learner) model for detection and classification. Using Bayesian optimization function to optimize the hyperparameters is performed to choose the best performance configuration. Through experimental analyses, the higher detection accuracy of the proposed method is 98.5%, and it is proved that the proposed model has good convergence, accuracy, stability and generalization ability.
- Open Access
- 14 October 2021
- Published by Springer Nature in BMC Medical Education
- Vol. 21 (1) , 1-9
An observation of the peer-assisted learning (PAL) method in the clinical teaching of vertigo/dizziness-related diseases for standardized residency training (SRT) students in China: a randomized, controlled, multicenter study
Background: Vertigo and dizziness (VD) are among the most frequently seen symptoms in clinics and are important for medical students, especially for those in Chinese standardized residency training (SRT). The aim of our study was to examine the PAL method’s feasibility in the clinical teaching of VD-related diseases for SRT students in China. Methods: This is a randomized, controlled, multicenter study. A total of 228 residents were invited to participate in this study, of which 198 completed the program. The students were randomized into two groups, and VD-related diseases were taught using lecture-based learning (control group) or peer-assisted learning (PAL). An examination paper and a rating scale were used to evaluate students’ performance in the mastery of VD-related theoretical knowledge and clinical skills, meanwhile students’ perceptions, satisfaction, and risk of burnout were also analyzed using a questionnaire. Independent-samples t-test and chi-square analysis were performed to evaluate statistical significance for continuous variables and categorical variables, respectively, using SPSS 18.0 software. Results: The PAL group performed better in mastering theoretical knowledge and clinical skills than the control group. And more students believed that PAL could help improve their personal qualities such as teamwork skills. However, more students reported that PAL increased the risk of burnout. Conclusions: PAL was a suitable and effective method in the clinical teaching of some specialized diseases, especially it was recommended for students who had gained initial knowledge and skills, such as Chinese SRT students. However, we should draw attention to the increased risk of burnout if PAL is intended to be widely used in clinical teaching. Trial registration: ISRCTN registry, ISRCTN53773239, 05/07/2021, retrospectively registered.
- 14 October 2021
- Published by Springer Nature in BMC Medical Education
- Vol. 21 (1) , 1-9
Open AccessAn observation of the peer-assisted learning (PAL) method in the clinical teaching of vertigo/dizziness-related diseases for standardized residency training (SRT) students in China: a randomized, controlled, multicenter study
Background: Vertigo and dizziness (VD) are among the most frequently seen symptoms in clinics and are important for medical students, especially for those in Chinese standardized residency training (SRT). The aim of our study was to examine the PAL method’s feasibility in the clinical teaching of VD-related diseases for SRT students in China. Methods: This is a randomized, controlled, multicenter study. A total of 228 residents were invited to participate in this study, of which 198 completed the program. The students were randomized into two groups, and VD-related diseases were taught using lecture-based learning (control group) or peer-assisted learning (PAL). An examination paper and a rating scale were used to evaluate students’ performance in the mastery of VD-related theoretical knowledge and clinical skills, meanwhile students’ perceptions, satisfaction, and risk of burnout were also analyzed using a questionnaire. Independent-samples t-test and chi-square analysis were performed to evaluate statistical significance for continuous variables and categorical variables, respectively, using SPSS 18.0 software. Results: The PAL group performed better in mastering theoretical knowledge and clinical skills than the control group. And more students believed that PAL could help improve their personal qualities such as teamwork skills. However, more students reported that PAL increased the risk of burnout. Conclusions: PAL was a suitable and effective method in the clinical teaching of some specialized diseases, especially it was recommended for students who had gained initial knowledge and skills, such as Chinese SRT students. However, we should draw attention to the increased risk of burnout if PAL is intended to be widely used in clinical teaching. Trial registration: ISRCTN registry, ISRCTN53773239, 05/07/2021, retrospectively registered.
- Open Access
- 18 September 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 172152-172165
Two-Stream Convolutional Neural Network Based on Gradient Image for Aluminum Profile Surface Defects Classification and Recognition
In this article, a novel two-stream convolutional neural network based on gradient image is performed to effectively classify and identify aluminum profiles defects for the first time. Recent feature fusion methods based on two-stream network prove promising performance for defects classification and recognition. In this article, we use data enhancement methods to obtain a large number of samples to prevent the over fitting phenomenon in deep learning. The image gradient is calculated with the Sobel operator, and normalized to transform the data between zero and one under the same dimension. We design a two-stream convolutional neural network model adopting Wavelet transform fusion strategy to realize feature fusion on the ReLU6 layer, which uses the original RGB image of aluminum profile and the gradient image corresponding to the original RGB image as inputs to extract features through two sub-networks and fuses features on a concatenate layer to be input into SVM classifier for classification and recognition. Using Bayesian Optimization function and computing the cross-validation classification error to optimize the hyperparameters to choose the best performance configuration is performed. A series of experimental data, which include accuracy and estimated generalized classification errors of single-stream and two-stream networks with different feature fusion strategies on different fusion layers, are conducted and show that the current model has good convergence, accuracy, stability and generalization. On this basis, this article also proposes a series of innovative methods for the future research of other defects.
- 18 September 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 172152-172165
Open AccessTwo-Stream Convolutional Neural Network Based on Gradient Image for Aluminum Profile Surface Defects Classification and Recognition
In this article, a novel two-stream convolutional neural network based on gradient image is performed to effectively classify and identify aluminum profiles defects for the first time. Recent feature fusion methods based on two-stream network prove promising performance for defects classification and recognition. In this article, we use data enhancement methods to obtain a large number of samples to prevent the over fitting phenomenon in deep learning. The image gradient is calculated with the Sobel operator, and normalized to transform the data between zero and one under the same dimension. We design a two-stream convolutional neural network model adopting Wavelet transform fusion strategy to realize feature fusion on the ReLU6 layer, which uses the original RGB image of aluminum profile and the gradient image corresponding to the original RGB image as inputs to extract features through two sub-networks and fuses features on a concatenate layer to be input into SVM classifier for classification and recognition. Using Bayesian Optimization function and computing the cross-validation classification error to optimize the hyperparameters to choose the best performance configuration is performed. A series of experimental data, which include accuracy and estimated generalized classification errors of single-stream and two-stream networks with different feature fusion strategies on different fusion layers, are conducted and show that the current model has good convergence, accuracy, stability and generalization. On this basis, this article also proposes a series of innovative methods for the future research of other defects.