Outline: lab-hygiene-sentinel Feasibility Analysis
- Context seed — Scene: a clinical lab bench where cross-contamination risk is invisible; what this doc assesses and what the reader will know at the end.
- What the system must do — Trace: technician picks up instrument → puts it down → system starts 30s watch window → wash or violation → alert + clip; optional: compliance history, shift reports.
- Where sensor fusion and CV add value — Named tasks: instrument pickup detection (RFID + CV fusion), hand-wash event detection (sensor + CV), person attribution (badge/wristband RFID or camera tracking), violation classification, evidence clip generation.
- Architecture — Layers: sensor edge (RFID readers, sink sensor, cameras) → event broker → state machine engine → violation classifier → alert dispatcher + clip store; Mermaid component diagram.
- What is hard — Named challenges: person attribution in multi-technician lab; on-metal RFID reliability per instrument type; CV and RFID fusion without double-counting; 30s window enforcement with clock drift; video clip retrieval latency; clinical compliance audit trail.
- Feasibility verdict — 7 dimensions: instrument detection, hand-wash detection, person attribution, real-time alerting, evidence retrieval, regulatory compliance, optional features.
- Why these outputs — Why? table for: in-lab alarm, supervisor push alert, video evidence clip, violation log, (optional) compliance history, (optional) shift report.
- Build order — Spine:
Sensor Edge → Event Broker → State Machine → Alert Dispatcher; Bulges: CV redundancy layer (off Sensor Edge), clip store (off Event Broker), compliance history (off violation log), shift reports (off compliance history).
- Open questions — 4 items: compliance obligation confirmation; person attribution method; existing camera infrastructure; phone app platform.