Xuefeng Feng

Xuefeng Feng

Auto-generated profile, based on publicly available information

Xuefeng Feng

Xinjiang Uygur Autonomous Region Research Institute of Measurement and Testing, Urumqi, China

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
12 June 2024 in IEEE Sensors Journal
Verified
Journal Article
Closed access
LLLi Liu
Li Liu
XFXuefeng Feng
Xuefeng Feng
FLFeng Li
Feng Li
QXQinglong Xian
Qinglong Xian
ZCZhendong Chen
Zhendong Chen
ZJZhenhong Jia
Zhenhong Jia
Due to the frequent occurrence of missed detections, false alarms, and the low accuracy of surface defect detection in industrial components, surface defect detection has always been a challenge in the industrial field. In this study, we propose a novel network structure based on YOLOv5s that achieves improved defect detection accuracy by improving the YOLOv5s network structure. First, a CoTNet transformer module is incorporated in the feature extraction process of the backbone network, replacing the original C3 network module. A global attention mechanism (GAM) is integrated along with the C3 module into the neck to enhance the feature learning ability of the model so that it focuses more on features. Last, the adaptive spatial feature fusion algorithm is used for prediction to increase the forecasting accuracy of the model. Finally, the improved algorithm is evaluated on a dataset containing NEU-DET steel surface defects. The experimental results show that the improved algorithm in this paper outperforms other current state-of-the-art algorithms in terms of on this dataset detection accuracy. In addition, we conduct additional experiments on the PKU-Market-PCB dataset, confirming that the proposed method yields improved detection accuracy.

21 June 2024
Verified
Conference Paper
Open access
ZCZhendong Chen
Zhendong Chen
XFXuefeng Feng
Xuefeng Feng
LLLi Liu
Li Liu
ZJZhenhong Jia
Zhenhong Jia
Early inspection of the blade surface defects of wind turbines not only reduces costly repairs, but also prevents the occurrence of grave defects in safety. The current objective inspection methods in detecting surface defects of wind turbine blades have the problems of insufficient inspection capability and high complexity of inspection model. Therefore, this paper carries out a relevant research related to the issues mentioned above, and put forward a mechanism for detecting surface defects on wind turbine blades, improving the detection accuracy while significantly diminishing the parameter number and computational volume of the algorithm model. First, the MobileNetV3 network is introduced and innovated by the YOLOv5, and the MobileNetV3-N network is proposed, which better improves the capability of the network to focus on and captivate the target to be detected; Then, the standard convolutional layers in the YOLOv5 algorithmic neck network are replaced with phantom convolutional modules to lighten the network model; Finally, an attention efficient mechanism is integrated into the neck network of the YOLOv5 algorithm to better balance the detected accuracies on the algorithm. Based on the homemade wind turbine blade surface defects dataset as well as the public dataset NEU-DET, several sets of comparative experiments and ablation experiments have been conducted using the innovative algorithm, and the results of the experiments have verified the advantages of the proposed algorithm.