Target identification method, target identification device, electronic device, and storage medium
By using a Block layer model in a real-time pedestrian and vehicle detection system, combined with partial relational attention mechanism and global contrastive pooling attention mechanism, multi-scale features are extracted and fused, solving the problem of insufficient feature extraction capability and achieving more accurate target recognition and tracking.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- DONGGUAN ZKTECO ELECTRONICS TECH
- Filing Date
- 2026-03-31
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, real-time detection and tracking systems for pedestrians and vehicles lack sufficient feature extraction capabilities, leading to inaccurate target recognition and affecting tracking performance.
The first model, which includes a Block layer, is used to extract features from video frames. The Block layer consists of a convolutional layer, a first attention layer, a second attention layer, and a first fusion layer. Multi-scale features are extracted and fused using a partial AND relational attention mechanism and a global contrastive pooling attention mechanism to generate discriminative feature vectors.
It improves the accuracy and robustness of target recognition, and enhances the ability to track and recognize targets in complex scenarios.
Smart Images
Figure CN122416331A_ABST