Chengjie Gu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Chengjie Gu has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Chao Pan
School of Computer Science and Information Engineering, Hefei University of Technology, Hefei, Anhui, China
1 shared publication - Chenxi Hu
School of Computer Science and Engineering, Anhui University of Science and Technology, Huainan, China
1 shared publication - Donghui Hu
Hefei University of Technology, Hefei, China
1 shared publication Dongjun Zhu
Anhui University of Science and Technology
1 shared publication- Gang Xu
Southeast University
1 shared publication - Hang Li
Shenzhen Research Institute of Big Data, Shenzhen, Guangdong, China
1 shared publication
Latest Publications
- 5 June 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (23) , 236001
MSDroneDet: a lightweight hypergraph-semantics guided fusion framework for multimodal remote sensing object detection
Single-modal remote sensing object detection often suffers from performance degradation in diverse scenarios. In contrast, multimodal remote sensing detection enhances target feature representation through cross-modal data fusion. Current mainstream deep learning methods adopt convolutional neural networks or Transformer-based fusion techniques for feature complementarity, but they can only capture low-order cross-modal dependencies and introduce high computational overhead. This limitation impedes cross-modal extraction and hinders the establishment of robust inter-modal target correlations. We propose MSDroneDet, an innovative cross-modal hypergraph semantics-guided fusion framework. It implements hierarchical high-level semantic modeling through the hypergraph semantic cross-modal module, explicitly encoding non-local spatial high-order dependencies across multimodal domains. We design a lightweight feature extraction bottleneck module (P-C3k2) based on partial convolution. It can effectively capture shallow-layer features and reduce computational complexity significantly while maintaining detection performance. Furthermore, we introduce an Inner-SIoU loss function that incorporates angular and scaling factors to regulate auxiliary bounding box dimensions, thereby enhancing the efficacy of regression. Evaluations conducted on two RGB-IR remote sensing benchmarks demonstrate that our approach achieves state-of-the-art detection accuracy with moderate computational overhead, validating its effectiveness for multimodal remote sensing object detection.
- 5 June 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (23) , 236001
MSDroneDet: a lightweight hypergraph-semantics guided fusion framework for multimodal remote sensing object detection
Single-modal remote sensing object detection often suffers from performance degradation in diverse scenarios. In contrast, multimodal remote sensing detection enhances target feature representation through cross-modal data fusion. Current mainstream deep learning methods adopt convolutional neural networks or Transformer-based fusion techniques for feature complementarity, but they can only capture low-order cross-modal dependencies and introduce high computational overhead. This limitation impedes cross-modal extraction and hinders the establishment of robust inter-modal target correlations. We propose MSDroneDet, an innovative cross-modal hypergraph semantics-guided fusion framework. It implements hierarchical high-level semantic modeling through the hypergraph semantic cross-modal module, explicitly encoding non-local spatial high-order dependencies across multimodal domains. We design a lightweight feature extraction bottleneck module (P-C3k2) based on partial convolution. It can effectively capture shallow-layer features and reduce computational complexity significantly while maintaining detection performance. Furthermore, we introduce an Inner-SIoU loss function that incorporates angular and scaling factors to regulate auxiliary bounding box dimensions, thereby enhancing the efficacy of regression. Evaluations conducted on two RGB-IR remote sensing benchmarks demonstrate that our approach achieves state-of-the-art detection accuracy with moderate computational overhead, validating its effectiveness for multimodal remote sensing object detection.
- 5 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23 (3) , 5000-5014
Generative Image Steganography With Minimum-Distance Guidance
Image steganography conceals secret data within a digital image while preserving its innocent appearance. The advent of artificial intelligence generative models has given rise to a new paradigm known as generative image steganography, which hides secret data directly into the image generation process. However, existing generative image steganographic methods are typically only applicable to unquantized stego images, severely limiting their practicality in real-world scenarios. To address this limitation, we propose a generative image steganography with minimum-distance guidance based on a diffusion model, called MDStega. During the hiding phase, MDStega designs a secret data-driven residual image sampling mechanism, which establishes a dynamic mapping relationship between discrete secret data and continuous probability distributions, strictly preserving the distribution consistency between stego images and normally generated images. During the extraction phase, the minimum-distance guidance rule effectively suppresses the interference caused by stego image quantization on the extraction accuracy of secret data. Furthermore, MDStega does not require fine-tuning pre-trained models or training additional models, which significantly reduces computational overhead and training time. Experimental results demonstrate that MDStega is superior to state-of-the-art methods by not only ensuring secure concealment at 3 bits per pixel (bpp) in PNG format but also achieving a recovery accuracy of up to 99%, demonstrating strong practical potential.
- 5 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23 (3) , 5000-5014
Generative Image Steganography With Minimum-Distance Guidance
Image steganography conceals secret data within a digital image while preserving its innocent appearance. The advent of artificial intelligence generative models has given rise to a new paradigm known as generative image steganography, which hides secret data directly into the image generation process. However, existing generative image steganographic methods are typically only applicable to unquantized stego images, severely limiting their practicality in real-world scenarios. To address this limitation, we propose a generative image steganography with minimum-distance guidance based on a diffusion model, called MDStega. During the hiding phase, MDStega designs a secret data-driven residual image sampling mechanism, which establishes a dynamic mapping relationship between discrete secret data and continuous probability distributions, strictly preserving the distribution consistency between stego images and normally generated images. During the extraction phase, the minimum-distance guidance rule effectively suppresses the interference caused by stego image quantization on the extraction accuracy of secret data. Furthermore, MDStega does not require fine-tuning pre-trained models or training additional models, which significantly reduces computational overhead and training time. Experimental results demonstrate that MDStega is superior to state-of-the-art methods by not only ensuring secure concealment at 3 bits per pixel (bpp) in PNG format but also achieving a recovery accuracy of up to 99%, demonstrating strong practical potential.
- 3 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network Science and Engineering
- Vol. 13, 6402-6417
Robust Power Allocation for UAV-Aided ISAC Systems With Uncertain Location Sensing Errors
Integrated sensing and communication (ISAC) enables simultaneous sensing and data transmission with the assistance of unmanned aerial vehicles (UAVs) in emergency disaster relief and inspects scenarios. However, the impact of sensing uncertainty on communication performance has not been systematically investigated. In this paper, we propose a novel UAV-aided ISAC framework that explicitly accounts for the uncertainty location sensing error (LSE). To characterize LSE more realistically, we derive the Cramér-Rao bound (CRB) and use it as the variance parameter for the considered uncertainty LSE models, instead of adopting the conventional unit-variance assumption. Then, we analytically reveal the inherent coupling relationship between LSE and achievable communication rate. Considering three practical LSE distributions, namely, ellipsoidal, Gaussian, and arbitrary distributions, we formulate three robust communication and sensing power allocation problems and develop tractable solutions using the ${\cal {S}}$-Procedure with alternating optimization (${\cal {S}}$-AO) method, Bernstein-type inequality with successive convex approximation (BI-SCA) method, and conditional value-at-risk (CVaR) with AO (CVaR-AO) method. Simulation results validate the theoretical coupling, demonstrate the robustness of the proposed schemes, and reveal sensing-communication trade-offs, providing valuable insights for robust UAV-aided ISAC system design.
- 3 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network Science and Engineering
- Vol. 13, 6402-6417
Robust Power Allocation for UAV-Aided ISAC Systems With Uncertain Location Sensing Errors
Integrated sensing and communication (ISAC) enables simultaneous sensing and data transmission with the assistance of unmanned aerial vehicles (UAVs) in emergency disaster relief and inspects scenarios. However, the impact of sensing uncertainty on communication performance has not been systematically investigated. In this paper, we propose a novel UAV-aided ISAC framework that explicitly accounts for the uncertainty location sensing error (LSE). To characterize LSE more realistically, we derive the Cramér-Rao bound (CRB) and use it as the variance parameter for the considered uncertainty LSE models, instead of adopting the conventional unit-variance assumption. Then, we analytically reveal the inherent coupling relationship between LSE and achievable communication rate. Considering three practical LSE distributions, namely, ellipsoidal, Gaussian, and arbitrary distributions, we formulate three robust communication and sensing power allocation problems and develop tractable solutions using the ${\cal {S}}$-Procedure with alternating optimization (${\cal {S}}$-AO) method, Bernstein-type inequality with successive convex approximation (BI-SCA) method, and conditional value-at-risk (CVaR) with AO (CVaR-AO) method. Simulation results validate the theoretical coupling, demonstrate the robustness of the proposed schemes, and reveal sensing-communication trade-offs, providing valuable insights for robust UAV-aided ISAC system design.
- 22 September 2025
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (15) , 1-31
CPA-TAM: channel patch aggregation and topological association mining for visible-infrared person re-identification
- 22 September 2025
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (15) , 1-31
CPA-TAM: channel patch aggregation and topological association mining for visible-infrared person re-identification