一种多模态对齐层级融合的星载遥感森林树种识别方法
By using a multimodal alignment hierarchical fusion network, the problems of feature misalignment and semantic gap in tree species identification in complex forest areas using multimodal remote sensing data are solved. It achieves accurate alignment and deep interactive fusion of multi-source information, improves the accuracy and robustness of tree species identification, and is applicable to forest resource surveys and management using spaceborne remote sensing data.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HEFEI UNIV OF TECH
- Filing Date
- 2026-04-27
- Publication Date
- 2026-07-17
AI Technical Summary
Existing multimodal remote sensing data fusion methods have failed to effectively address the issues of feature misalignment, information distortion, and semantic gap in tree species identification in complex forest areas, making it difficult to fully leverage the complementary advantages of multimodal information. In particular, they suffer from problems such as small coverage and high cost in spaceborne lidar applications.
A multimodal alignment hierarchical fusion network is adopted. Differential features are extracted from spaceborne multispectral, lidar and high-resolution images through a three-branch feature extraction module. Cross-modal accurate alignment and hierarchical fusion are achieved by using an intramodal constraint module and a cross-modal guided attention mechanism. Feature interaction fusion is performed by combining scale-adaptive deformable convolution and hybrid attention blocks.
It significantly enhances the collaborative expression capability of multi-source remote sensing information in complex forest backgrounds, improves the accuracy and robustness of tree species identification at the regional scale, and strengthens the ability to distinguish tree species with similar spectra. It has good regional adaptability and promotion and application value.
Smart Images
Figure CN122090290B_ABST