Personnel Detection
Personnel detection is an AI recognition algorithm that detects and locates the full body of a person in images or video. It serves as the foundational technology for various security and safety management algorithms such as perimeter intrusion detection, boundary violation detection, crowd detection, loitering detection, and fall detection.
Algorithm Overview
Using deep learning-based object detection methods, this algorithm detects person positions in images with bounding boxes. It supports full-body detection and provides stable detection even in diverse postures such as standing, sitting, and lying down.
Performance Metrics
| Dataset | Detection Accuracy (mAP@0.5) |
|---|---|
| PERSON | 0.79 |
Edge AI Board (RV1126B) Execution Efficiency
| Algorithm | Processing Time |
|---|---|
| person_detect | 71ms |
Key Features
- Full-body detection support: Handles diverse postures including standing, sitting, and lying down
- Wide-area monitoring: Simultaneous detection of multiple people with a single camera
- Foundation for derivative algorithms: Used as a pre-processing stage for intrusion detection, loitering detection, fall detection, and more
- Long-term stable operation: Supports continuous day and night operation
Use Cases
- Factory and warehouse safety zone monitoring (detection of entry into hazardous areas)
- Construction site personnel deployment confirmation
- Occupancy and bed-leaving detection in nursing care facilities
- Office attendance management
- Store congestion measurement
- Public facility dwell detection and loitering detection
Edge AI Board Implementation
Utilizing the RV1126B NPU, personnel detection is executed with an inference time of 71ms. Camera video is processed in real time, and detection results can be notified via network.