Dengfeng Chen

Dr. Dengfeng Chen


Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

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

Representative Publications

View All Publications
1 January 2026 in Engineering Research Express
Verified
Journal Article
Closed access
ZZZhonglin Zhang
Zhonglin Zhang
DCDengfeng Chen
Dengfeng Chen
JYJingyao Yang
Jingyao Yang
JLJiahao Li
Jiahao Li
HXHaiyan Xiao
Haiyan Xiao
QLQianqian Liu
Qianqian Liu
Photovoltaic panels are susceptible to environmental conditions and manufacturing defects during operation. Timely automated detection is crucial for maintaining power generation efficiency and ensuring equipment safety. This paper presents a lightweight enhanced YOLOv11n model for detecting photovoltaic panel defects through visible–infrared image fusion. The proposed model integrates a cross-modulation fusion network that extracts structural texture features from visible images and thermal anomaly features from infrared images, enabling effective cross-modal fusion without explicit geometric registration. Moreover, to enhance both detection accuracy and adaptability for edge deployment, the model incorporates a Semantic-Aware Downsampling Module (NAMDown), a Lightweight Feature Extraction Module (LFEM), and a Location Quality Estimation Head (LQE-Head). Furthermore, the InterpIoU loss function, tailored for small targets, is employed to enhance regression performance. Experiments are conducted on a self-constructed dual-modal dataset for photovoltaic panel defects. The results indicate that the proposed model achieves 81.33% accuracy and surpasses existing lightweight detection methods in both accuracy and recall. Finally, the model is deployed on the Jetson AGX Orin edge platform. With TensorRT optimization, the average inference delay is reduced to 29.4 ms, demonstrating a favorable balance between accuracy and real-time performance.

29 March 2024 in Sensors
Verified
Journal Article
Open access
PDF file
FLFeng Dong Lei
Feng Dong Lei
LZLiang Zhao
Liang Zhao
DCDengfeng Chen
Dengfeng Chen
Classifying the flow subsequences of sensor networks is an effective way for fault detection in the Industrial Internet of Things (IIoT). Traditional fault detection algorithms identify exceptions by a single abnormal dataset and do not pay attention to the factors such as electromagnetic interference, network delay, sensor sample delay, and so on. This paper focuses on fault detection by continuous abnormal points. We proposed a fault detection algorithm within the module of sequence state generated by unsupervised learning (SSGBUL) and the module of integrated encoding sequence classification (IESC). Firstly, we built a network module based on unsupervised learning to encode the flow sequence of the different network cards in the IIoT gateway, and then combined the multiple code sequences into one integrated sequence. Next, we classified the integrated sequence by comparing the integrated sequence with the encoding fault type. The results obtained from the three IIoT datasets of a sewage treatment plant show that the accuracy of the SSGBUL–IESC algorithm exceeds 90% with subsequence length 10, which is significantly higher than the accuracies of the dynamic time warping (DTW) algorithm and the time series forest (TSF) algorithm. The proposed algorithm reaches the classification requirements for fault detection for the IIoT.

15 May 2023 in Sensors
Verified
Journal Article
Open access
PDF file
NWNingke Wei
Ningke Wei
LZLiang Zhao
Liang Zhao
DCDengfeng Chen
Dengfeng Chen
An enhanced self-assembling network routing algorithm is proposed for the problem of weak connectivity of communication networks caused by factors such as movement or environmental interference in the construction and operation, and the maintenance of construction robot clusters. Firstly, the dynamic forwarding probability is calculated based on the contribution of nodes joining routing paths to network connectivity, and the robust connectivity of the network is achieved by introducing the connectivity feedback mechanism; secondly, the influence of link quality evaluation index Q balanced hop count, residual energy, and load on link stability is used to select appropriate neighbors for nodes as the subsequent hop nodes; finally, the dynamic characteristics of nodes are combined with the topology control technology to eliminate low-quality links and optimize the topology by link maintenance time prediction and to set robot node priority. The simulation results show that the proposed algorithm can guarantee a network connectivity rate above 97% under heavy load, reduce the end-to-end delay, and improve the network survival time, providing a theoretical basis for achieving stable and reliable interconnection between building robot nodes.

19 October 2026
Verified
Dataset
Open access
NSNuo Su
Nuo Su
DCDengfeng Chen
Dengfeng Chen
ZZZhonglin Zhang
Zhonglin Zhang
HLHaoming Li
Haoming Li
This replication package accompanies the manuscript “CF-Drive: Candidate-conditioned interaction risk estimation for retrieval-augmented autonomous driving,” submitted to the Journal of Intelligent and Connected Vehicles under manuscript ID JICV-2026-0117. The package contains the source code, experimental configurations, log-disjoint fold definitions, processed candidate-level results, evaluation and plotting scripts, and a Replication Explanatory File for reproducing the principal risk-estimation, matched-shuffle, and utility-selection results. The original NAVSIM data are not redistributed. Users should obtain the source data from the official NAVSIM repository and comply with its applicable terms of use.

1 March 2024 in Preprints.org
Verified
Preprint
Open access
PDF file
FLFeng Dong Lei
Feng Dong Lei
LZLiang Zhao
Liang Zhao
DCDengfeng Chen
Dengfeng Chen
Classifying network flow subsequences in an Industrial IoT gateway is an effective method for identifying Industrial IoT faults. This paper proposes a subsequence classification algorithm SSGBUL-IKNN based on unsupervised learning encoding to address the problem of low classification accuracy. Firstly, we built a network module based on unsupervised learning to encode the flow sequence of the Industrial IoT gateway. In the encoding module, a subsequence calibration algorithm controls the increasing error during the coding process. Then, the specific fault type is obtained through an improved KNN classification algorithm based on the indication distance source from the integrated coding sequences converted by multi-dimensional network flow sequences. The test result on three Industrial IoT datasets of a sewage treatment plant shows that the accuracy of SSGBUL-IKNN in this paper exceeds 92%, significantly higher than the algorithm DTW and TSF. It reaches the classification requirements of Industrial IoT fault diagnosis.

9 March 2026 in Engineering Research Express
Verified
Journal Article
Closed access
LZLiang Zhao
Liang Zhao
JLJiahao Li
Jiahao Li
ZSZhan Ping Song
Zhan Ping Song
DCDengfeng Chen
Dengfeng Chen
ZZZhonglin Zhang
Zhonglin Zhang
XLXu Li
Xu Li
As a critical defect affecting the safety and durability of tunnel structures, detection of joint fissures plays a vital role in ensuring tunnel construction safety.This study proposes a joint detection model that integrates attention and super-resolution mechanisms into YOLOv8n, specifically designed for complex tunnel environments. Based on YOLOv8n, a lightweight multi-scale residual attention (MSRA) module is designed to enhance the representational ability of target features and strengthen the semantic expression of feature information. It also achieves a balance between lightweight design and multiscale feature extraction capability. Meanwhile, to improve model diversity and discrimination while reducing computational overhead, a lightweight multi-scale pooling module (DSAP), which combines multi-scale pooling, dual-channel attention, and feature recalibration, is introduced. In addition, to ensure efficient utilization of feature information, a combination of an efficient attention mechanism, a progressive upsampling strategy, and a residual learning module is designed. This component, termed the progressive super-resolution (APSR) module, is designed for image restoration tasks. Finally, experiments were conducted on a self-built dataset. The results demonstrate that the proposed model outperforms other mainstream models in terms of precision and recall, achieving a balance between reducing computational overhead and improving detection accuracy. The mean average precision (mAP) of the proposed model in the generalization experiment reaches 75.6%, exceeding that of other models, thereby verifying its superior robustness.