JHJinglu He

Jinglu He

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Jinglu He has a total of 41 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • MSLAN: multi-scale spatial location awareness network for crowd counting

    by Na Li, Jiaqi Chen, Xingxuan Jia, Shuyi Yan, Jiaxuan Zhang, Jinglu He

  • Open Access

    Global-Local-Structure Collaborative Approach for Cross-Domain Reference-Based Image Super-Resolution

    by Xiuxia Cai, Chenyang Diwu, Ting Fan, Wenjing Wang, Jinglu He

    Remote sensing image super-resolution (RSISR) aims to reconstruct high-resolution images from low-resolution observations of remote sensing data to enhance the visual quality and usability of remote sensors. Real world RSISR is challenging owing to the diverse degradations like blur, noise, compression, and atmospheric distortions. We propose hierarchical multi-task super- resolution framework including degradation-aware modeling, dual-decoder reconstruction, and static regularization-guided generation. Speciffcally, the degradation-wise module adaptively characterizes multiple types of degradation and provides effective conditional priors for reconstruction. The dual-decoder platform incorporates both convolutional and Transformer branches to match local detail preservation as well as global structural consistency. Moreover, the static regularizing guided generation introduces prior constraints such as total variation and gradient consistency to improve robustness to varying degradation levels. Extensive experiments on two public remote sensing datasets show that our method achieves performance that is robust against varying degradation conditions.

  • Open Access

    Global-Local-Structure Collaborative Approach for Cross-Domain Reference-Based Image Super-Resolution

    by Xiuxia Cai, Chenyang Diwu, Ting Fan, Wenjing Wang, Jinglu He

    Remote sensing image super-resolution (RSISR) aims to reconstruct high-resolution images from low-resolution observations of remote sensing data to enhance the visual quality and usability of remote sensors. Real world RSISR is challenging owing to the diverse degradations like blur, noise, compression, and atmospheric distortions. We propose hierarchical multi-task super- resolution framework including degradation-aware modeling, dual-decoder reconstruction, and static regularization-guided generation. Speciffcally, the degradation-wise module adaptively characterizes multiple types of degradation and provides effective conditional priors for reconstruction. The dual-doder platform incorporates both convolutional and Transformer branches to match local detail preservation as well as global structural consistency. Moreover, the static regularizing guided generation introduces prior constraints such as total variation and gradient consistency to improve robustness to varying degradation levels. Extensive experiments on two public remote sensing datasets show that our method achieves performance that is robust against varying degradation conditions.

  • Hierarchical multi-modal feature fusion for RGBT tracking

    by Na Li, Kai Huang, Zihang Wang, Yuquan Gan, Jinglu He


  • Open Access

    Texture Interaction Complement for Reference-based Image Super-Resolution

    by Chaojun Zhang, Yinghua Li, Ying Zhang, Jinglu He, Dan Xu

  • Open Access

    CPINet: Towards A Novel Cross-Polarimetric Interaction Network for Dual-Polarized SAR Ship Classification

    by Jinglu He, Ruiting Sun, Yingying Kong, Wenlong Chang, Chenglu Sun, Gaige Chen, Yinghua Li, Zhe Meng, Fuping Wang

    With the rapid development of the modern world, it is imperative to achieve effective and efficient monitoring for territories of interest, especially for the broad ocean area. For surveillance of ship targets at sea, a common and powerful approach is to take advantage of satellite synthetic aperture radar (SAR) systems. Currently, using satellite SAR images for ship classification is a challenging issue due to complex sea situations and the imaging variances of ships. Fortunately, the emergence of advanced satellite SAR sensors has shed much light on the SAR ship automatic target recognition (ATR) task, e.g., utilizing dual-polarization (dual-pol) information to boost the performance of SAR ship classification. Therefore, in this paper we have developed a novel cross-polarimetric interaction network (CPINet) to explore the abundant polarization information of dual-pol SAR images with the help of deep learning strategies, leading to an effective solution for high-performance ship classification. First, we establish a novel multiscale deep feature extraction framework to fully mine the characteristics of dual-pol SAR images in a coarse-to-fine manner. Second, to further leverage the complementary information of dual-pol SAR images, we propose a mixed-order squeeze–excitation (MO-SE) attention mechanism, in which the first- and second-order statistics of the deep features from one single-polarized SAR image are extracted to guide the learning of another polarized one. Then, the intermediate multiscale fused and MO-SE augmented dual-polarized deep feature maps are respectively aggregated by the factorized bilinear coding (FBC) pooling method. Meanwhile, the last multiscale fused deep feature maps for each single-polarized SAR image are also individually aggregated by the FBC. Finally, four kinds of highly discriminative deep representations are obtained for loss computation and category prediction. For better network training, the gradient normalization (GradNorm) method for multitask networks is extended to adaptively balance the contribution of each loss component. Extensive experiments on the three- and five-category dual-pol SAR ship classification dataset collected from the open and free OpenSARShip database demonstrate the superiority and robustness of CPINet compared with state-of-the-art methods for the dual-polarized SAR ship classification task.


  • Open Access

    Light-weight Super-Resolution Network based on Classified Measurement-domain Features

    by Shuo Li, Yinghua Li, Dan Xu, Ying Zhang, Jinglu He

    In contemporary research, natural images are frequently reconstructed using methods that are computationally intensive but highly effective. To solve this issue, we propose a light-weight super-resolution(SR) network based on classified measurement-domain features, abbreviated as SRCMF. Images of nature can be divided into multiple smaller images for processing purposes. On the basis of this line of reasoning, it is possible to conclude that various image regions present unique recovery challenges and can be processed by models with differing degrees of complexity. Textured regions make supersegmentation more difficult than flat regions. To take advantage of this property, we can manage the various subgraphs of the split using an appropriate SR model, i.e. a simple small model for smooth regions and a complex large model for textured regions. With this as our starting point, we propose the SRCMF, a classification and supersegmentation integrating framework. Specifically, it employs Class Modules to categorize subgraphs based on their recovery difficulty, followed by numerous SRModules to super-score the subgraphs in each category. The computational cost can be drastically reduced because the preponderance of sub-images will travel through fewer networks. ClassSR is founded on a compression-aware image block processing method on the measurement domain, while SRModule is comprised of SR models of varying proportions. When reconstructing natural images, experimental data indicate that our SRCMF can save FLOPs and reduce the computational complexity of reconstruction. This general structure is applicable to additional simple visual duties.

  • Open Access

    Hyperspectral Anomaly Detection based on Autoencoder using Superpixel Manifold Constraint

    by Yuquan Gan, Wenqiang Li, Ying Liu, Jinglu He, Ji Zhang

    In the field of hyperspectral anomaly detection, autoencoder (AE) have become a hot research topic due to their unsupervised characteristics and powerful feature extraction capability. However, autoencoders do not keep the spatial structure information of the original data well during the training process, and is affected by anomalies, resulting in poor detection performance. To address these problems, a hyperspectral anomaly detection method based on autoencoders with superpixel manifold constraints is proposed. Firstly, superpixel segmentation technique is used to obtain the superpixels of the hyperspectral image, and then the manifold learning method is used to learn the embedded manifold that based on the superpixels. Secondly, the learned manifold constraints are embedded in the autoencoder to learn the potential representation, which can maintain the consistency of the local spatial and geometric structure of the hyperspectral images (HSI). Finally, anomalies are detected by computing reconstruction errors of the autoencoder. Extensive experiments are conducted on three datasets, and the experimental results show that the proposed method has better detection performance than other hyperspectral anomaly detectors.