Safe Behaviour & Pose Analysis
Catch risky posture, climbing and reaching into machinery with skeletal analysis.
Most incidents come not from missing equipment but from habits: reaching into a machine, climbing a railing, lifting a load with the back.
Pose analysis extracts 17 body keypoints, making posture measurable. “Walking past a machine” and “reaching into a machine” become distinguishable, and ergonomic risk is managed as a trend rather than as isolated incidents.
How it works
Posture extracted
Keypoints are computed per person; works independently of clothing and lighting.
Interpreted over time
Consecutive frames are combined to classify bending, reaching and climbing.
Calibrated to your site
The angle and duration that count as “risky” are set on your site during the pilot.
At which station and at what time of day does risky movement increase?
Fatigue is the strongest indicator of ergonomic risk: risky bending rises markedly in the last two hours of a shift. Seeing this tells you the answer is not training but rotation, work-height adjustment or lifting equipment.
| Risky bending (back) | 412 |
| Reaching into machinery | 188 |
| Climbing a railing | 96 |
| Running / abrupt motion | 74 |
| One-handed load carrying | 61 |
Back bending accounts for more than half of all events — a pattern that can be reduced directly with a palletizing table set to a fixed height.
| h1 | 18 |
| h2 | 21 |
| h3 | 24 |
| h4 | 22 |
| h5 | 29 |
| h6 | 34 |
| h7 | 39 |
| h8 | 43 |
The fatigue curve differs per site; when break planning follows this chart, event counts drop.
Use cases
Press and die
Stop signal the moment someone reaches into the machine.
Manual handling
Detection of back-lifting patterns.
Work at height
Railing-crossing and climbing alerts.
Ergonomics report
Monthly risk score per station.
Technical requirements
| Camera | Side or oblique angle preferred; person height ≥ 120 px in frame |
| Lighting | Works under variable lighting; no backlight source in frame |
| Processing | The pose model requires a GPU; 6–10 cameras per server |
| Output | Event records, ergonomics score, station and shift breakdown |
| Data | No identity matching; the track ID is deleted when the person leaves the frame |
Frequently asked questions
How does Safe Behaviour & Pose Analysis work?
Most incidents come not from missing equipment but from habits: reaching into a machine, climbing a railing, lifting a load with the back.
Are our existing cameras suitable for this solution?
Required camera specification: Side or oblique angle preferred; person height ≥ 120 px in frame. Before installation we assess your camera inventory free of charge and report which points are suitable and, if needed, which camera should be added.
What results and indicators does it provide?
Key indicators: Keypoints in the posture model: 17; Real-time analysis: 25 fps; Defined risky-behaviour classes: 4. Measurement method: pose analysis. Back bending accounts for more than half of all events — a pattern that can be reduced directly with a palletizing table set to a fixed height. Thresholds are calibrated on your own site during the pilot.
Is it compliant with KVKK, Türkiye's data-protection law?
Footage is processed on an edge server inside your facility; raw video is not sent to the cloud. Facial recognition is not used; analytics outputs are anonymous and event clips are deleted automatically when the defined retention period ends.