System Overview
- AI surveillance tool
- Crowd surveillance
- Rule-based event detection
- People attribution recognition
- Identification and human behavior recognition
- Machine learning algorithm
- Video surveillance search engine
Crowd Surveillance
- People counting
- Line counting – crossing a specific point
- Pass counting – entering/leaving a specific point
- 88% accuracy
- 95% accuracy after re-configuration
Rules-Based Detection
- Person intrusion
- Person tailgating
- Loitering
- Crowd alert
- Reverse movement
- Abandoned item
People Attribution Identification
- Gender (male/female)
- Long/short hair
- Wearing/not wearing glasses
- Wearing/not wearing hat
- Wearing/not wearing t-shirt
- Wearing long/short sleeves
- Wearing/not wearing jeans
- Wearing/not wearing shorts
- Wearing long/short pants
- Wearing clothes with pattern/without pattern
- Other customized attributes (optional)
- Maximum accuracy over 90%
Human Behavior Recognition
- Person falling
- Waving
- Banging head
- Kicking towards wall/door
- Fighting
- Other customizable actions
Video Search Engine
- Natural language
- Real-time video search engine
- Algorithmic deep learning tools
- Every nanosecond of video searchable
- Recognizes objects, colors, people, vehicles, animals and more
Technical Features
- Support video input from IP camera, USB camera, and video file.
- Support scheduling of analysis
- Automatic detection and alert of camera tampering and video quality deterioration
- Supports real-time mosaic insertion over person for privacy protection
- Capable of operating in adverse environments such as low light, rain, snow and clouds
System Applications
- Crowd monitoring at events
- Security surveillance at premise for automatic intrusion detection and alarm
- Queue monitoring
- Crowd flow monitoring at transport terminals (airport, bus-stop, metro station)
- Abnormality detection and alarm in public space, such as loitering, abandoned item, fighting, crowd gathering
