一种红外遥感图像轻量化小目标识别方法、设备及介质
By constructing a lightweight backbone network HLNet and a multi-scale feature fusion network LSF, and combining depthwise separable convolution and Transformer architecture, the accuracy and robustness issues of small target recognition in infrared remote sensing images are solved, achieving efficient and real-time small target recognition in complex backgrounds.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING INSTITUTE OF TECHNOLOGY (ZHUHAI)
- Filing Date
- 2026-04-21
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies for small target recognition in infrared remote sensing images suffer from insufficient detection accuracy and robustness, making it difficult to balance the requirements of high precision, lightweight design, and high efficiency. In particular, in complex backgrounds, small target feature extraction is insufficient, and recognition stability and generalization ability are inadequate.
We construct a lightweight backbone network HLNet and a lightweight multi-scale feature fusion network LSF, and combine deep separable convolution and Transformer architecture to achieve efficient recognition of small infrared targets through multi-scale feature fusion and global semantic modeling.
While maintaining the model's lightweight design, it significantly improves the accuracy and stability of infrared small target recognition in complex backgrounds, making it suitable for nighttime monitoring, shadow area recognition, and real-time target detection under low-light conditions.
Smart Images
Figure CN122090041B_ABST