Inspection Detection Coverage Index (IDCI)

GMI-TEL-038TelecommunicationsInfrastructure-DependentPattern O-7: Inspection Coverage

Definition

Ratio of defect types detectable by the drone/CV inspection system to the full set of defect types that a comprehensive physical inspection is expected to identify, measured by defect category. IDCI is diagnostic of the CV model's detection breadth: its output is not just a score but a list of uncovered defect types that defines the mandatory supplementary physical inspection scope.

The Measurement Gap

Drone inspection programs report total defects found but do not measure which defect categories the CV model is structurally blind to. An inspection that finds 200 defects across 8 categories but misses 4 entire defect categories creates a false sense of thoroughness: the count is high, but the coverage is incomplete. Without a systematic enumeration of detectable vs. undetectable defect types, organizations cannot determine whether drone inspection is a complete replacement for climber inspection or a partial supplement requiring mandatory physical follow-up for specific defect categories.

The Decision

Which defect categories is the drone/CV system blind to, and for those categories, is a supplementary physical inspection required to maintain coverage? The answer directly shapes the inspection program design: if IDCI shows 80% coverage, the supplementary physical inspection scope is defined by the specific 20% of defect types the CV model cannot detect. If IDCI shows 50% coverage, drone inspection is a supplement, not a replacement, and physical climber inspections remain the primary method for half the inspection standard.

Research Foundation

Aerial drone inspection systems demonstrate high detection rates for surface-visible defects (corrosion, structural deformation, missing hardware) but consistently fail to detect subsurface conditions, internal corrosion, foundation degradation, and defects requiring physical interaction such as torque verification (TIA-222 Structural Standard for Antenna Supporting Structures, 2024). FCC tower registration inspection requirements encompass defect categories that include both surface-visible and physical-interaction items (47 CFR Part 17: Construction, Marking, and Lighting of Antenna Structures).

Formula

IDCI = Defect_types_with_demonstrated_CV_detection / Total_defect_types_in_inspection_standard

Per-category binary:
IDCI_c = 1 if CV true positive rate for defect type c exceeds detection threshold, 0 otherwise

Variable Definitions (every variable resolved)

  • Defect_types_with_demonstrated_CV_detection = defect categories where the CV model achieves true positive rate >= 0.7 on a validation dataset of known defects (count)
  • Total_defect_types_in_inspection_standard = complete set of defect categories in the applicable inspection checklist (count)
  • Detection threshold = 0.7 true positive rate (proposed starting point: lower than typical ML thresholds because the consequence of a missed structural defect is severe, so even modest detection capability is valuable compared to zero)

Units:

Ratio, 0 to 1. 1.0 = the CV model can detect every defect type in the standard. 0.5 = half the defect types are outside the model's demonstrated capability.

Edge Cases

Defect types inherently undetectable from aerial imagery (internal corrosion, foundation soil conditions, subsurface anchor degradation): classify as "structurally undetectable by drone" and exclude from the denominator; report separately as mandatory supplementary inspection scope. Defect types the model has never been tested against: classify as "untested" rather than "undetectable". They require validation, not permanent exclusion. Model updates may expand detection capability over time: re-evaluate IDCI after every model version update. Evaluate against the complete inspection standard for each asset type (tower, pole, cabinet, manhole).

These are proposed starting points: calibrate to your operational context during pilot: IDCI above 0.8 means the drone/CV system covers the large majority of the inspection standard. Below 0.6 means physical inspection remains required for more than 40% of defect types. Any defect type with failure consequence rated safety-critical and IDCI_c = 0 must be covered by mandatory supplementary physical inspection at the applicable regulatory frequency.

Confidence Grading (required alongside every IDCI value):

Grade Basis Interpretation
A: Empirical 5+ observed events in measurement window provide direct tail observation Measurement grounded in observed outcomes; highest confidence
B: Hybrid 1-4 observed events plus Monte Carlo simulation with Bayesian updating Blends observation and simulation; moderate confidence
C: Simulated Zero observed events; derived entirely from Monte Carlo simulation Modeled, not observed; treat as indicative, not validated

A IDCI of any value at Grade C is a simulation that agrees with itself. It tells you the model is internally consistent but not that it matches reality. Grade C scores should trigger proactive stress-testing against historical near-miss events and cross-asset analogues rather than acceptance as calibrated.

Proposed formula, pending validation pilots.

At a Glance

Unit
Ratio (0 to 1)
Trend Direction
Higher
Reporting Frequency
Per model update
Category
AI Inspection Accuracy and Regulatory Sufficiency
← Defect Severity Grading Accuracy (DSGA)Inspection Skill Retention Index (ISRI) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment