Success case
People Flow Analysis in Public Space – Especially Avoid Missing Detection due to Occlusion
Hardware: NVIDIA Jetson Orin NX Application: People Flow Analysis for Dense Crowd Industry: Public Safety Deployment Location: Europe
- Industry
- Public Safety & Security
- Application
- Public-space people counting
Project background and needs
This people-flow solution targets European public spaces such as malls, stations, and subways, using Jetson Orin NX to process crowd count, direction, and trajectory information.
Dense occlusion
- In high-density crowds, individuals occlude each other.
- Occlusion can cause missed and spurious identifications.
Needs effective detection in dense scenes.
Continuous identity tracking
- Traditional detection struggles to identify pedestrian direction and identity.
- Re-identifying pedestrians after temporary occlusion, plus distractions like mannequins, affects the count.
Needs to link trajectories across frames and distinguish real pedestrians.
Solution and architecture
Detection and re-identification
FairMOT combines CenterNet detection with ReID features, linking targets across frames to avoid double-counting the same person.
Density-based training
The solution uses whole-body detection for sparse scenes and head tracking for dense scenes; both run on Jetson Orin NX.
Solution components
On-site gallery (3)
Discuss your application with our team
Tell us about your requirements and deployment environment.