AGENTIC AI & ENGINEERING · 2026-10-08 · 2 MIN READ

Drisva Architecture: Real-Time Industrial Surveillance and Computer-Vision Anomaly Detection

How Drisva processes existing RTSP/ONVIF facility camera feeds on local edge appliances to detect safety hazards, perimeter intrusions, and process anomalies.

1. The Limits of Passive CCTV Monitoring Walls

Most manufacturing plants, warehouses, hospitals, and commercial campuses already have dozens of IP cameras installed across loading bays, shop floors, and perimeters. Yet in nearly every facility, those cameras function only as passive recording devices—reviewed hours or days after an incident has already occurred because no human operator can continuously watch 32 or 64 video feeds at once.

At the same time, industrial and institutional operators are rightly reluctant to stream live internal camera feeds to external cloud servers—both because of uplink bandwidth constraints and because facility video is sensitive operational data. We built Drisva, Astrake’s dedicated industrial surveillance and anomaly detection system, to turn existing camera networks into local, proactive alerting systems.

2. Local RTSP/ONVIF Edge Inference Pipeline

Drisva connects directly to your existing NVR or IP cameras over the local facility LAN using standard RTSP and ONVIF streams—requiring no proprietary camera replacements:

  • Hardware-Decoded Frame Sampling: A local Drisva edge inference node ingests multi-camera streams on-site, decoding frames in hardware and running computer-vision detection pipelines tailored to each camera’s zone and schedule.
  • Zone & Schedule Rules: Each camera view can be configured with polygonal regions of interest (such as high-voltage switchgear zones, forklift aisles, or after-hours loading docks) and shift-based active rules.
  • Temporal Debouncing: Rather than triggering noisy alerts on a single motion frame, Drisva verifies persistence across consecutive frames before logging an event and notifying supervisors.

3. Industrial Safety, Perimeter & Process Anomaly Detection

Drisva focuses on practical visual scenarios across industrial and commercial environments:

TABLE 01 // DRISVA COMPUTER VISION DETECTION MODULES
Detection Category Monitored Conditions Supervisor Alert Output
Shop-Floor Safety & PPE Hard-hat, high-visibility vest, and restricted machinery-zone entry checks Timestamped keyframe clip, camera zone tag, and local console notification
Perimeter & After-Hours Security Person or vehicle entry into gated yards, warehouses, or restricted corridors outside shift hours Immediate alert to security desk with option to trigger zone audio warnings via Arenea
Process & Facility Anomalies Conveyor blockages, aisle obstructions, or early visual smoke/spill indicators Exception log routed to shift supervisors for rapid human verification

4. Local Video Custody, DPDP Readiness & Human-in-the-Loop Review

Because Drisva processes video streams on on-premise edge appliances inside your facility network, raw camera footage never needs to leave your premises. Event logs are stored with configurable retention schedules and role-based access in technical alignment with India DPDP Act 2023 privacy principles, keeping every automated alert subject to human supervisor verification.

Explore our Artificial Intelligence & Vision Systems domain or contact our engineering team to evaluate Drisva for your facility.

Want to Discuss This Topic With Our Team?

Reach out to explore how we can apply these principles to your organization.

START A CONVERSATION