基于近红外光谱与可见光图像的多模态检测柑橘患虫方法
By fusing visible light images and near-infrared spectral features using a multimodal network model, the problems of low efficiency and misjudgment in traditional citrus pest detection are solved, achieving accurate and non-destructive detection of citrus pests and improving detection accuracy and reliability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUAZHONG AGRI UNIV
- Filing Date
- 2025-05-15
- Publication Date
- 2026-07-17
AI Technical Summary
Traditional methods for detecting citrus pests are inefficient and cannot achieve non-destructive testing. Single near-infrared spectroscopy detection is prone to missing latent pests, and pure visual detection is difficult to distinguish between mechanical damage and pest characteristics, leading to misjudgment.
By combining multi-angle visible light images and near-infrared spectral data, and fusing image features and spectral features through a multimodal network model, including image processing branches, spectral processing branches, cross-modal self-attention mapping mechanism, feature stitching and classification layers, pest detection can be achieved.
It improves the accuracy and reliability of citrus pest detection, ensures a non-damaging detection process, and provides pest control and quality assurance for the citrus industry.
Smart Images
Figure CN120544234B_ABST
Abstract
Citation Information
Patent Citations
Unmanned aerial vehicle small target detection method based on multispectral interactive attention fusion
CN117830878A
Cross-modal pedestrian re-identification method based on bimodal alignment
CN117935307A