Spore-pollen image recognition method and system based on multi-modal dialogue large model
By constructing a pollen image recognition system based on a multimodal dialogue model, the problems of traditional pollen identification relying on expert experience and the lack of interactive capabilities in single-modal models are solved. This system achieves efficient and accurate automatic pollen identification and analysis, and possesses professional interactive capabilities.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HENAN UNIVERSITY
- Filing Date
- 2026-03-27
- Publication Date
- 2026-07-17
AI Technical Summary
Traditional pollen identification methods rely heavily on expert experience, resulting in low efficiency. Furthermore, existing single-modal deep learning models struggle to integrate multi-source information and lack interactive capabilities, leading to insufficient accuracy and efficiency in pollen identification.
A pollen image recognition method based on a multimodal dialogue model is adopted. Through a multi-stage image enhancement process, a hierarchical label smoothing strategy, a pollen recognition rule base, and domain-specific prompt word engineering, combined with efficient parameter fine-tuning technology, a GLM4v multimodal dialogue model is constructed to achieve automatic recognition and analysis of pollen images.
It significantly improves the accuracy and efficiency of pollen identification, especially the fine-grained classification accuracy of families, genera, and species, and has natural language interaction capabilities, providing professional morphological identification basis and reasoning process, thus enhancing the interpretability and practicality of the results.
Smart Images
Figure CN122415453A_ABST