Prism — Detect
See what's happening. Know what it means. Act before it escalates.
Computer vision models running continuously against your camera streams — detecting asset conditions, safety events, and compliance breaches as they occur, structuring findings automatically, and routing outputs through the path your operations require. One camera can run multiple models simultaneously, generating different insights for different teams from the same stream.
Hero: live camera frame with CV detection overlay (bounding box, condition label, severity badge) alongside structured detection record panel.
Continuous CV Analysis
Purpose-built computer vision models run continuously across your connected camera streams. Models are trained for specific asset classes, environments, and event types. One camera stream can run multiple models — asset condition monitoring and PPE compliance and perimeter detection — each routing to the appropriate team and output path.
Detection pipeline: camera stream → multiple CV models in parallel → different findings routing to different output paths and stakeholders.
Safety & compliance detection
- PPE compliance — hard hat, hi-vis vest, harness, safety glasses
- Proximity and exclusion zone breaches — people or vehicles in restricted areas
- Perimeter intrusion detection — unauthorised access alerts in real time
- Vehicle safe zone and tracking — real-time position relative to defined zones
- Hazard identification — damaged equipment, leaks, spills, unsafe conditions
- Incident response support — real-time footage analysis for evacuations, near-misses, collisions
Asset condition detection
- Crusher bridging, conveyor belt tracking, oversize fragmentation, overheating rollers
- Equipment anomaly and condition change — structural, operational, visual deviation
- Material build-up, transfer chute deviation, filter press and OBF condition
- Smoke, flare, and emissions monitoring
- Missing GET / tooth wear, bucket load, safety zone proximity
- Continuous analysis across all connected camera streams, multiple models per stream
Detection Records
Every detection structured: condition type, severity, confidence score, camera source, timestamp, and full asset or area linkage — in a single record. Immediately reviewable, filterable, and ready for the appropriate output path.
- Structured detection record per identified condition
- Severity classification and prioritization
- Camera frame reference and source footage link per detection
- Asset and geospatial linkage
- Filterable by severity, condition type, asset, site, and time window
Detection records list: Condition Type, Asset/Area, Site, Severity, Confidence, Timestamp, Output Path, Status.
Output Path — Analytics
When detections indicate a condition to track, investigate, or report on, Prism routes findings into the analytics layer — structured data linked to footage, surfaced in condition trend dashboards and exportable reports.
- Detection frequency and severity trend charts per asset or area
- Condition change tracking over time, linked to source footage
- Exportable condition and compliance reports (PDF, CSV, JSON)
- Detection histories per asset for maintenance planning and audit
- Portfolio-level condition comparison across assets and sites
Output Path — Alerts
When detections meet defined conditions or cross defined thresholds, Prism routes the finding as a real-time alert — to the right people or directly to operational systems.
- Real-time notifications to stakeholders via email, SMS, and mobile push
- Integration alert output to SCADA, ERP, and work order platforms
- Severity-based escalation and routing rules
- Alert acknowledgement, assignment, and resolution tracking
- Maintenance window and suppression rule management
Alert routing diagram: detection finding → condition threshold check → two output paths: notifications to stakeholders, and integration trigger to SCADA, ERP, or work order system.
Review, Annotation & Model Development
Review, annotate, and escalate detection findings from within Prism — confirm classifications, add diagnostic context, attach evidence, and link to work orders. Build, train, and deploy custom computer vision models against your specific camera environments, asset types, and event definitions. Models improve with your data.
- In-platform annotation and reclassification
- Escalation and work order linkage
- Custom model development within Prism, trained on your labeled footage
- Performance validation before deployment
- One-click deployment, version control, and rollback
- Shared model library for common industrial asset classes and environments
How it fits into the Prism workflow
AI Detection & Alerts processes the continuous camera streams visible in Live Monitoring and produces structured findings that feed directly into the analytics output or alert routing — closing the loop from camera to action.
Workflow position
- Capture
- Monitor
- Detect
- Act
Outcomes
- Fewer unplanned failures — asset conditions detected visually before they escalate to downtime
- Faster incident response — safety events surfaced and routed immediately to the right people or systems
- Stronger compliance posture — every detection timestamped, classified, and linked to footage; audit evidence on demand
- Real-time situational awareness — operations and safety teams working from the same live picture, from any location
- Lower false positive rate — models trained for your specific environments and event types outperform generic alerting
