HONGJIE LI

Dr. HONGJIE LI

orcid

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

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

Representative Publications

View All Publications
12 May 2025 in Sensors
Verified
Journal Article
Open access
PDF file
HLHONGJIE LI
HONGJIE LI
YLYang Liu
Yang Liu
CLChuni Liu
Chuni Liu
HPHongxuan Pang
Hongxuan Pang
Ke XuKe Xu
Ke Xu
Few-shot steel surface defect generation remains challenging due to the limited availability of training samples and the complex visual characteristics of industrial defects. Traditional data augmentation techniques often fail to capture the diverse manifestations of defects, resulting in suboptimal detection performance. While existing stable diffusion models possess robust generative capabilities, they encounter domain knowledge transfer limitations when applied to industrial settings. To address these constraints, stable industrial defect generation (stableIDG) is proposed, which enhances stable diffusion through three key improvements: (1) fine-tuning with low-rank embedding adaptation to accommodate steel surface defect generation; (2) implementation of personalized identifiers associated with defect regions to prevent malformation during generation; and (3) development of a mask-guided defect learning mechanism that directs network attention toward critical defect regions, thereby enhancing the fidelity of generated details. The method was validated on a newly constructed Medium and Heavy Plate Surface Defect Dataset (MHPSD) from a real industrial environment. Experiments demonstrated that stableIDG achieved better image generation quality compared to existing methods, such as textual inversion. Furthermore, when the generated samples were used for data augmentation, the detection network performance was effectively enhanced. Detection recall significantly improved for both defect classes, increasing from 0.656 to 0.908 for Inclusion defects and from 0.86 to 0.986 for Foreign Object Embedding defects. These results demonstrate the effectiveness of the proposed method for industrial defect detection in data-scarce scenarios.

3 February 2026 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
XCXiangyu Cao
Xiangyu Cao
SWShun Wang
Shun Wang
HLHONGJIE LI
HONGJIE LI
Dongdong ZhouDongdong Zhou
Dongdong Zhou
Ke XuKe Xu
Ke Xu
To address the technical challenges in online surface defects detection of hot continuous casting slab (HCCS), this paper presents the design, development, and validation of a novel vision-based instrumentation system based on Multi-View Multispectral Photometric Stereo (MVMSPS). The system integrates four cameras with specific spectral illuminators to intrinsically eliminate spectral crosstalk at inspection speeds up to 0.5 m/s. The core innovations are twofold: 1) A Generalized Homography Transformation (GHT) method is proposed for precise multi-view image matching, reducing geometric alignment errors by over 87.8% compared to conventional methods. 2) The optimal illumination zenith angle under the Lambertian model is theoretically deduced to guide the lighting design and maximize the signal-to-noise ratio. Enabled by these innovations, the MVMSPS algorithm effectively fuses multi-view and multispectral data to effectively estimate the surface normal and capture the high-frequency information of surface depth variations. Experimental results demonstrate that the proposed MVMSPS system can accurately, reliably, and efficiently detect surface defects (e.g., scars, cracks, and indentations) on HCCS in high-temperature and dynamic environments, with a complete processing time of under 3.6s per image set, satisfying the real-time requirements for online inspection and offering a promising instrument solution for intelligent production and quality control in the steel industry. Additionally, this paper explores potential future research directions, including combining fast two-dimensional detection with three-dimensional photometric stereo and integrating multimodal information, to further enhance detection performance.

10 January 2026 in Applied Soft Computing
Verified
Journal Article
Closed access
QSQian Sun
Qian Sun
Ke XuKe Xu
Ke Xu
HLHONGJIE LI
HONGJIE LI
CLChuni Liu
Chuni Liu
LJLei Jin
Lei Jin
DZDelong Zhao
Delong Zhao

23 September 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
ZSZhuangyi Shi
Zhuangyi Shi
HLHONGJIE LI
HONGJIE LI
QTQiusheng Tan
Qiusheng Tan
ZJZeyu Jia
Zeyu Jia
BABopei An
Bopei An
SLShanglin Lv
Shanglin Lv
Alloy steel wire rods are extensively utilized in critical industries, including automotive manufacturing, machinery production, bridge cables, energy equipment, fasteners, steel cords, bearing steel, and spring steel. The microstructural uniformity, mechanical properties, and dimensional stability of these wire rods directly influence the quality of subsequent processing and their service reliability. In wire rod production, phase transformation control in the air-cooling section post-laying is vital; thus, high-precision online temperature detection is essential for ensuring the quality of the wire rods. Currently, temperature detection following the laying process primarily relies on point-based or local-area measurement methods. Although some studies have explored infrared area-array cameras for online temperature measurement and thermal field analysis, existing methods fail to fulfill practical requirements for full-field temperature detection, temperature drop tracking, and precise analysis of same-coil temperature differences in the air-cooling section. These shortcomings stem from limitations in field-of-view coverage, the complex coiled wire rod structure, and challenges in accurately identifying overlap points. To address the issues regarding insufficient temperature measurement accuracy, challenges in effective overlap point identification, and the quantitative analysis of same-coil temperature differences during online temperature detection of air-cooled wire rods, this paper presents a systematic investigation. First, a four-channel multispectral image sensor integrating near-infrared, far-infrared, and visible light channels was developed, followed by system integration for full-field online temperature detection and laboratory calibration using a blackbody furnace. Subsequently, deep learning-based methods for image enhancement, head/tail recognition, overlap point identification, and a two-color thermometry model were proposed. The developed temperature detection system was implemented in a high-speed wire rod production line at a steel plant in China, where actual on-site data were gathered, and temperature calibration was performed using a portable blackbody cavity and point thermometers. This research evaluated temperature drop tracking from the laying head exit through the phase transformation zone to subsequent cooling regions and conducted an analysis of same-coil temperature differences at various wire rod positions. The findings demonstrate that this research not only achieves high-precision, full-field temperature detection and drop tracking for air-cooled wire rods but also facilitates overlap point localization and quantitative analysis of same-coil temperature differences. This work provides valuable data to support phase transformation zone monitoring, optimization of cooling regimes, and quality control in the air-cooled wire rod production process.

18 June 2026 in Applied Soft Computing
Verified
Journal Article
Closed access
HLHONGJIE LI
HONGJIE LI
CLChuni Liu
Chuni Liu
QSQian Sun
Qian Sun
XCXiangyu Cao
Xiangyu Cao
Ke XuKe Xu
Ke Xu

13 July 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
HLHONGJIE LI
HONGJIE LI
JTJingya Tao
Jingya Tao
DZDongdong Zhou
Dongdong Zhou
SLShanglin Lv
Shanglin Lv
The in-situ detection of near-surface defects (NSDs) in complex metallic components, particularly within confined geometries such as nuclear steam turbine heat sinks, is significantly hindered by spatial limitations and harsh environmental noise. Traditional non-destructive testing (NDT) methodologies frequently fall short in sensitivity to micro-defects and lack the mechanical flexibility required for such confined inspections. This study presents an innovative integrated laser-based active thermography system specifically designed to overcome these challenges in restricted environments. The system features a compact fiber-coupled semiconductor laser, effectively decoupling the bulky excitation source from the inspection head to enhance accessibility to narrow gaps. Utilizing 3D finite element thermal conduction simulations, we optimized the excitation parameters (90 W, 10 s) to maximize thermal contrast while preventing surface ablation. To ensure robust defect characterization under low signal-to-noise ratio (SNR) conditions, a hierarchical data processing pipeline was developed: (1) A Transformer-based Restormer network is employed for image restoration, achieving an 8.9% improvement in SNR compared to Gaussian filtering by effectively suppressing background noise while preserving critical high-frequency defect edges; (2) A quantitative inversion model, which integrates Fuzzy C-Means (FCM) clustering with a Back Propagation (BP) neural network, is established, yielding width estimation errors within 5% and depth prediction errors predominantly within 0.06 mm; (3) An enhanced YOLOv11 architecture, augmented with adaptive noise filtering and spatial attention mechanisms, is deployed for automated defect recognition. Experimental validation on standard steel specimens and authentic nuclear heat sinks demonstrated a mean Average Precision (mAP@50) of 97.9%, underscoring the system’s capacity to deliver high-precision, quantitative, and automated solutions for industrial inspections in confined spaces.