SurveillX intelligence

From continuous activity to an answer you can verify.

SurveillX learns site rhythms at the edge, selects the moments that carry information, and turns them into searchable, cited operational evidence—without pretending that a sampled frame knows more than it does.

A higher-signal operating record

Do more than detect an object.

The intelligence path combines deployed Panoptic services with Cognia on sites configured for extended edge analytics.

Learn

Understand the rhythm.

Activity is accumulated into time windows and compared with learned hour-of-week baselines, so unusual volume and duty cycle can surface after the model has enough history.

Select

Keep the frames that matter.

Alerts, anomalies, visual novelty, and periodic baselines create a compact visual history instead of treating every frame as equally informative.

Retrieve

Search meaning and appearance.

Panoptic combines caption and summary meaning with native image embeddings to find relevant events, activity windows, and visual evidence.

Verify

Answer with citations.

Structured answers return supporting summary, image, and event identifiers, confidence, and uncertainty—not just fluent text.

Explain

Measure activity, not impressions.

Quantitative analysis calculates peaks, averages, active windows, and per-class patterns directly from stored activity data.

Report

Build a grounded field narrative.

Daily, weekly, and arbitrary-window reports combine activity metrics, selected evidence, events, coverage, and configured site context.

Intelligence architecture

Observe locally. Understand across time. Return to the source.

Raw video remains part of the evidence path. Intelligence records point operators toward the right time, camera, event, and frame.

Cognia edge analytics

Learn + select

  • Detection + tracking
  • ROI and tripwire rules
  • Activity buckets + baselines
  • Anomaly + scene novelty
  • Coverage-aware edge events
Panoptic intelligence

Describe · search · verify

  • Caption + summarize
  • Text and image embeddings
  • Hybrid evidence retrieval
  • Quantitative analysis
  • Grounded answers + reports
SurveillX operations

Review + act

  • Customer-scoped access
  • Ask and search workflows
  • Saved report definitions
  • Evidence image retrieval
  • Source-frame pull when needed
Field cameras observing a changing remote site
Cognia-configured edgeActivity · anomaly · novelty

Cognia edge intelligence

Learn what normal looks like before declaring what changed.

On configured sites, Cognia turns continuous detections into activity windows, evaluates scheduled rules, learns weekly operating patterns, and selects high-signal images for the connected intelligence record.

  • FOV, ROI, and tripwire rules with duration, clear, severity, and cooldown behavior
  • Fifteen-minute activity windows and learned hour-of-week baselines
  • Warm-up-aware anomaly scoring across event count and active duty cycle
  • Visual novelty plus periodic baseline frames for a representative site history
  • Scheduled alert gates and temporary overrides that expire predictably

Separate deployment · model profile confirmed per site

Panoptic search + Ask

Ask the site. Inspect the evidence behind the answer.

Authorized operators can search within customer, site, camera, time, trigger, and event boundaries.

Evidence-grounded inquiry

Questions route to the right evidence path.

Panoptic can choose quantitative analysis, evidence Q&A, activity intensity, visual progress, or an on-demand source frame. The response preserves provenance and can decline or reframe a question when the record cannot support it.

SITE INTELLIGENCE / EVIDENCE QUERY
When did activity start this morning, and what evidence supports that?
Scope customer, site, cameras, and local timescope
Measure the first active windows by detection classSQL
Retrieve nearby summaries, events, and imageshybrid
Check coverage and cite supporting identifiersverify
The answer includes a time value, confidence, cited evidence IDs, and uncertainty where camera coverage or sampled imagery limits the conclusion.

Operational analysis

Separate what was measured from what was inferred.

Activity metrics come from structured windows. Visual descriptions come from selected images. Coverage states tell the operator how complete that view was.

Activity intensityPeak and average levels, active windows, per-object statistics, hourly distribution, event markers, and coverage-aware narrative.
Visual progressChronological baseline, novelty, alert, and anomaly images describe how selected views evolved across cameras.
Unified event layerAlerts, anomalies, scene changes, activity spikes, after-hours activity, activity drops, starts, and late starts share a searchable model.
Source-frame returnWhen selected evidence is not enough, the system can request a fresh or forensic frame from Continuum at a bounded time.
Three coverage statesActivity, observed-empty, and unknown/offline prevent missing camera data from being misreported as a quiet site.
Structured verificationClaims can be labeled supported, partially supported, not supported, or insufficient evidence, with confidence and uncertainties.
Operator reviewing visual evidence and a structured field timeline
Report evidence linkedMetrics · events · selected images

Grounded reporting

Reports that know the site, the window, and the limits.

Daily, weekly, and custom-window reports can be configured around cameras, detection classes, operating hours, content sections, local display time, and the narrative appropriate to the site. On sensor-equipped sites, recorded weather and sensor state sit beside the footage as reviewable context for what the report describes.

  • Saved customer-scoped report definitions and asynchronous run history
  • Site context, operating schedule, cameras, object classes, and selected sections
  • Activity statistics plus semantic evidence and visual-progress summaries
  • Coverage and evidence references carried with the generated result

Claim boundaries

Useful intelligence requires honest limits.

No. Cognia is a separately deployed edge analytics stack. Camera support, rules, model profile, compute capacity, schedules, and event policies are confirmed for the target site.

No. Panoptic creates a searchable intelligence record from activity windows and selected evidence. Continuum remains the source archive, and operators can return to nearby footage or request a source frame when needed.

No. A sampled frame supports what was visible at that moment. Activity windows support measured detection activity, while coverage data states whether the camera observed the full period. The system keeps those evidence types distinct.

No. Safety and PPE models are not presented as validated compliance guarantees. Any such use requires separate model validation, operating procedures, human review, and contractual scope.

Bring a real question

Test the intelligence against your field workflow.

Choose a site question, reporting window, camera set, and evidence standard. We’ll show which answer path is current, which requires Cognia, and where the system should say it does not know.