一种基于机器视觉的多模式目标检测方法及系统
By using image grayscale conversion, edge recognition, and graph neural network to generate an attention field, the problems of poor flexibility and insufficient lighting adaptability of existing target detection systems in embedded scenes are solved, achieving high-precision and stable target detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- APPLIED TECH COLLEGE OF SOOCHOW UNIV
- Filing Date
- 2025-12-08
- Publication Date
- 2026-07-17
AI Technical Summary
Existing target detection systems are difficult to deploy in cost-sensitive or space-constrained embedded scenarios, have poor flexibility, low distance calculation accuracy, weak anti-interference ability, poor light adaptability, and traditional methods are sensitive to changes in light, resulting in a decrease in recognition accuracy.
A multi-mode target detection method combining image grayscale, edge recognition, and graph neural networks is adopted. By generating initial and final attention fields, it dynamically adapts to changes in illumination, enhances anti-interference ability, improves distance calculation accuracy, and extracts geometric information through neural networks.
It improves the system's flexibility and recognition accuracy, enhances the stability of target detection in complex lighting environments, significantly improves distance calculation accuracy and anti-interference capability, and adapts to different shapes and task requirements.
Smart Images
Figure CN121661330B_ABST