一种企业多模态图像智能识别处理系统
By generating multimodal image association maps and dynamically adjusting feature weights, the problems of cross-stage tracing errors and the inability to directly translate recognition results into business actions in multimodal image recognition technology are solved, achieving high-precision intelligent recognition of enterprise multimodal images and system optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- QINGDAO QIANYUAN JIUZE ELECTRONIC INFORMATION TECHNOLOGY CO LTD
- Filing Date
- 2025-11-04
- Publication Date
- 2026-07-17
AI Technical Summary
Existing multimodal image recognition technologies cannot establish dynamic correlations, resulting in high error rates when tracing across links. They cannot directly convert recognition results into executable business actions, and the system cannot receive business execution feedback for model optimization, thus preventing continuous improvement in recognition accuracy.
The system employs an association mapping module to generate multimodal image association maps, corrects sample weight biases through a spatiotemporal weight scene correction algorithm, dynamically adjusts association feature weights using a three-level architecture combined with the feature extraction module, transforms recognition results into business suggestions through an identification and decision suggestion module, and encrypts and stores the data through a data security protection module, presents the entire process information using a visualization module, and performs iterative optimization using an optimization module.
It achieves accurate correlation between multimodal images, reduces tracing errors, supports cross-stage analysis, generates actionable business suggestions, reduces manual disassembly steps, ensures model optimization with business iterations, and improves recognition accuracy and system security.
Smart Images
Figure CN121543789B_ABST