A millimeter wave security door target recognition model training method
By extracting features and transforming dimensions from millimeter-wave security gate echo data, and using the Softmax function to calculate predicted probabilities and update model parameters, the accuracy of target recognition is improved, solving the problem that high-dimensional features are difficult to capture fine-grained features.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING HUAHANG RADIO MEASUREMENT & RES INST
- Filing Date
- 2024-12-30
- Publication Date
- 2026-06-09
AI Technical Summary
In millimeter-wave security gate echo target recognition tasks, high-dimensional features are difficult to capture the fine-grained characteristics of the target, resulting in low target classification and recognition accuracy.
By extracting features and transforming dimensions from batches of raw millimeter-wave security gate echo data, feature maps for each target category are obtained. Vector aggregation is then performed on the feature maps for each target category, and the prediction probability is calculated using the Softmax function to update the model parameters until the training termination condition is met.
It improves the accuracy of target recognition and classification, fully extracts the characteristics of high-dimensional features, and solves the problem that high-dimensional features are difficult to capture fine-grained features of targets.
Smart Images

Figure CN122174034A_ABST