Security AI · Human movement analysis
Moving / Stopped Persons Detection
An open-source computer vision system for real-time and post-event human movement analysis, detecting people and distinguishing between moving, stopped, and newly appeared individuals.
The challenge
Detect people in dynamic scenes
Identify individuals across diverse environments in real time or from recorded video.
Separate movement states
Distinguish between people who are moving, stopped, or newly detected in the frame.
Support security use cases
Flag stationary individuals whose prolonged stillness may require operator attention.
The approach
Person detection with YOLO
Used Ultralytics YOLOv8 / YOLOv10 to detect people across video frames.
Tracking and movement analysis
Used Roboflow Supervision to analyze displacement and classify motion behavior.
Absolute and relative thresholds
Implemented both absolute speed and relative speed threshold variants.
Scenario-based outputs
Separated combined moving/stopped detection, moving-only, and stopped-only workflows.
Outcomes
Movement classification
Moving, stopped, and under-processing states.
Scenario implementation
Absolute, relative, moving-only, and stopped-only workflows.
Stationary behavior monitoring
Designed for use cases where prolonged stillness may need attention.