Vehicle Detection
Vehicle detection is an AI recognition algorithm that identifies the position and type of vehicles in images or video. It serves as the foundational technology for traffic management and parking management, including illegal parking detection, congestion detection, vehicle counting, and vehicle type classification.
Algorithm Overview
Using deep learning-based object detection methods, vehicles are detected from parking lot and road video. It supports diverse vehicle types including passenger cars, trucks, buses, and motorcycles, and provides stable detection even at night and in adverse weather conditions.
Performance Metrics
| Dataset | Detection Accuracy (mAP@0.5) |
|---|---|
| CAR | 0.78 |
Edge AI Board (RV1126B) Execution Efficiency
| Algorithm | Processing Time |
|---|---|
| car_detect | 70ms |
Key Features
- Multi-vehicle type support: Detection of passenger cars, trucks, buses, and motorcycles
- Day and night operation: Stable detection at night when combined with night vision cameras
- Wide-area monitoring: Simultaneous detection of multiple vehicles with a single camera
- Foundation for derivative functions: Pre-processing for illegal parking, congestion, vehicle counting, etc.
Use Cases
- Parking lot occupancy management (vehicle counting)
- Illegal parking and stopping detection
- Traffic volume survey and congestion detection
- Vehicle intrusion detection (pedestrian zones)
- Logistics warehouse vehicle entry/exit management
- EV charging station utilization monitoring
Edge AI Board Implementation
Using the RV1126B NPU, vehicle detection is executed at 70ms. Combined with outdoor cameras, autonomous vehicle management is realized at the edge.