Jinshi Liu

Prof. Jinshi Liu

Shenzhen University, Shenzhen, Guangdong, China
orcid

Research Focus & Metrics

Representative Publications

View All Publications
1 January 2025 in IEEE Transactions on Image Processing
Verified
Journal Article
Closed access
LHLei He
Lei He
ZYZunhui Yi
Zunhui Yi
JLJinshi Liu
Jinshi Liu
CCChao-Yang Chen
Chao-Yang Chen
MLMing Lu
Ming Lu
ZCZhipeng Chen
Zhipeng Chen
The absorption and scattering of light in different turbid media cause images to suffer from poor visibility and contrast, which severely affects the performance of many computer vision tasks. To address this issue, we propose a fast scene recovery method based on the Ambient light similarity prior (ALSP). In this method, the ambient light similarity metric is designed from both magnitude and orientation, which is embedded into the optical imaging model, and the estimation of scene transmission is derived by simplification and approximation. The estimation of the transmission map is very simple, and its time complexity is O(N), where N is the size of the input image. Moreover, we propose a progressive manner to determine the ambient light for both the near and far regions separately, which can effectively improve the brightness and color saturation of the restored image. Experiments performed in different scenes demonstrate that our method outperforms several state-of-the-art competitors in terms of efficiency and scene recovery performance.

8 November 2022 in IEEE Transactions on Industrial Electronics
Verified
Journal Article
Closed access
JLJinshi Liu
Jinshi Liu
Zhaohui JiangZhaohui Jiang
Zhaohui Jiang
WGWeihua Gui
Weihua Gui
Zhiwen ChenZhiwen Chen
Zhiwen Chen
Particle size detection (PSD) is used to obtain the particle size distribution of materials in the blast furnace charging process, which is significant for optimizing the gas flow distribution and ensuring stable production. However, due to the complex surface texture of the materials and the uneven illumination of the production environment, existing methods have difficulty obtaining the particle size distribution efficiently. This article proposes an end-to-end PSD system based on image segmentation to obtain the particle size distribution online with high accuracy. First, to further enhance the expression of edge features and reduce the interference of complex textures, an RGB-laser particle segmentation network (RLPNet) is developed to obtain high-precision segmentation images by camera-LiDAR sensor fusion. Moreover, to improve the fusion of RGB and laser features, a feature dual-recalibration module was designed and embedded in RLPNet, consisting of independent recalibration and joint recalibration with T-convolution. Finally, to reduce the error caused by missing edge particle pixels, an edge-recognition-based particle size calculation strategy (ERP) is presented. Experimental results demonstrate that the proposed method performs well on the constructed dataset and in industrial applications. With the segmentation accuracy of RLPNet reaching 64.19 $\%$ , the similarity between the particle size distribution predicted with ERP and the actual distribution reaches 79.19 $\%$ .

12 February 2024 in IEEE/CAA Journal of Automatica Sinica
Verified
Journal Article
Closed access
Zhaohui JiangZhaohui Jiang
Zhaohui Jiang
CXChuan Xu
Chuan Xu
JLJinshi Liu
Jinshi Liu
WLWeichao Luo
Weichao Luo
Zhiwen ChenZhiwen Chen
Zhiwen Chen
WGWeihua Gui
Weihua Gui
The copper disc casting machine is core equipment for producing copper anode plates in the copper metallurgy industry. The copper disc casting machine casting package motion curve (CPMC) is significant for precise casting and efficient production. However, the lack of exact casting modeling and real-time simulation information severely restricts dynamic CPMC optimization. To this end, a liquid copper droplet model describes the casting package copper flow pattern in the casting process. Furthermore, a CPMC optimization model is proposed for the first time. On top of this, a digital twin dual closed-loop self-optimization application framework (DT-DCS) is constructed for optimizing the copper disc casting process to achieve self-optimization of the CPMC and closed-loop feedback of manufacturing information during the casting process. Finally, a case study is carried out based on the proposed methods in the industrial field.

19 December 2022 in IEEE Transactions on Circuits and Systems for Video Technology
Verified
Journal Article
Closed access
JLJinshi Liu
Jinshi Liu
Zhaohui JiangZhaohui Jiang
Zhaohui Jiang
TCTing Cao
Ting Cao
Zhiwen ChenZhiwen Chen
Zhiwen Chen
CZChaobo Zhang
Chaobo Zhang
WGWeihua Gui
Weihua Gui
Unlike general image segmentation, highly complex particle images have significant challenges in labeling and segmentation due to the information occlusion and texture disturbance. Aiming at the highly under-segmentation problem caused by complex particle image segmentation, this paper proposes a Semi-supervised Hybrid-training Particle Segmentation framework (SHPS) based on skeleton-guided pseudo-labels. First, a pre-trained model is obtained by training a popular segmentation algorithm on partially labeled data. Then, a Background Skeleton-guided Pseudo-label generation algorithm (BSP) is proposed to generate pseudo-labels closer to the ground truth in terms of structural integrity based on coarse segmentation. The final segmentation model is obtained by training a mixed dataset consisting of labeled data and pseudo-labels from another partition on the pre-trained model. The skeleton differences of pseudo-labels and coarse segmentation are added to the loss function. Experimental results show that our method achieves 84.4% accuracy on mIoU with uniform label data distribution, which is 2.1% higher than the accuracy of UNet and reduces the degree of under-segmentation.

19 October 2025
Verified
Conference Paper
Closed access
PLPan Liu
Pan Liu
JLJinshi Liu
Jinshi Liu
While significant advances exist in pseudo-label generation for semi-supervised semantic segmentation, pseudolabel selection remains understudied. Existing methods typically use fixed confidence thresholds to retain highconfidence predictions as pseudo-labels. However, these methods cannot cope with network overconfidence tendency, where correct and incorrect predictions overlap significantly in high-confidence regions, making separation challenging and amplifying model cognitive bias. Meanwhile, the direct discarding of low-confidence predictions disrupts spatial-semantic continuity, causing critical context loss. We propose Confidence Separable Learning (CSL) to address these limitations. CSL formulates pseudo-label selection as a convex optimization problem within the confidence distribution feature space, establishing sample-specific decision boundaries to distinguish reliable from unreliable predictions. Additionally, CSL introduces random masking of reliable pixels to guide the network in learning contextual relationships from lowreliability regions, thereby mitigating the adverse effects of discarding uncertain predictions. Extensive experimental results on the Pascal, Cityscapes, and COCO benchmarks show that CSL performs favorably against state-of-the-art methods. Code and model weights are available at: https://github.com/PanLiuCSU/CSL.

16 November 2023 in Engineering Applications of Artificial Intelligence
Verified
Journal Article
Closed access
ZJZhaohui Jiang
Zhaohui Jiang
JLJinshi Liu
Jinshi Liu
ZCZhiwen Chen
Zhiwen Chen
WLWeichao Luo
Weichao Luo
CZChaobo Zhang
Chaobo Zhang
WGWeihua Gui
Weihua Gui