7 February 2024 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
The landslide is a widespread and devastating natural disaster, posing serious threats to human life, security, and natural assets. Investigating efficient methods for accurate landslide detection with remote sensing images has important academic and practical implications. In this article, we proposed an LS-YOLO, a novel and effective model for landslide detection with remote sensing images. We first built a multi-scale landslide dataset (MSLD) and introduced random seeds in the data augmentation to increase data robustness. Considering the multi-scale characteristic of landslides in remote sensing images, a multi-scale feature extraction module is designed based on Efficient Channel Attention, Average Pooling, and Spatial Separable Convolution. To increase the receptive field of the model, dilated convolution is employed to the decoupled head. Specifically, the context enhancement module consisting of dilation convolutions is added to the decoupled head regression task branch, and then the improved decoupled head is to replace the coupled head in YOLOv5s. Extensive experiments show that our proposed model has high performance for multi-scale landslide detection, and outperforms other object detection models (Faster RCNN, SSD, EfficientDet-D0, YOLOv5s, YOLOv7, and YOLOX). Compared to the baseline model YOLOv5s, the AP of the LS-YOLO for detecting landslides has increased by 2.18% to 97.06%. The code and MSLD will be available at https://github.com/wenjieo/LS-YOLO.
25 September 2023 in IEEE Geoscience and Remote Sensing Letters
Ship detection is of great importance in territorial security and marine environmental protection. However, the accurate detection of small ships is a challenging task in complex environments mainly due to small ships having few features on remote sensing images. In this letter, we propose a small ship detection model based on YOLOv5s. The major features of the proposed model include: (1) a detection layer is added with a shallow feature map in the Head and skip connections are employed in the Neck to improve the detection accuracy of small ships; (2) a novel and effective Hybrid Spatial Pyramid Pooling (HSPP) is proposed to fuse the local and global information of feature maps; (3) a Coordinate Attention mechanism is employed in the Backbone to augment the representations of small ships, and EIOU is used as the loss function for bounding box regression to enhance the localization accuracy of the proposed model; (4) K-means++ algorithm is used to obtain more reasonable anchors for small ship detection. We introduce the multi-scale ship dataset OSSD, which contains 10133 images. Experiments on LEVIR-Ship and OSSD validate the effectiveness of our proposed model. The code and OSSD will be available at https://github.com/wenjieo/Improved-YOLOv5s-for-small-ship-detection.
23 December 2025 in IEEE Transactions on Aerospace and Electronic Systems
To improve the accuracy and robustness of anti-UAV detection, this study introduces TransAUAV, a Transformer-based RGB- infrared image fusion detection network. This approach strengthens object feature representation by leveraging multi-modal data fusion and self-attention mechanisms. Specifically, 1) we propose a multi-modal attention fusion module, which enhances the complementarity of RGB and infrared image features through a dual-path attention mechanism; 2) we propose a cross-layer multi-scale Transformer module, which improves detection performance by extracting multi-scale features and facilitating cross-modal information interaction; 3) we propose a texture information focus module, which enhances the representation of local texture details. Additionally, this paper designs a hybrid loss function to improve feature discrimination capability and training efficiency. Experimental results on anti-UAV and Drone-detection datasets show that TransAUAV outperforms state-of-the-art methods on all evaluation metrics.
30 April 2026 in Chaos, Solitons & Fractals