70% of utility networks run alongside public roads. Yet most operators walk the line once a year — or hire expensive helicopters and drones. What if every single service trip was also an inspection?
1. The Problem: How Do We Inspect Networks Today?
Assessing the condition of overhead networks — poles, conductors, insulators — is critical, but costly and time-consuming. Most operators currently rely on the following methods:
| Method | Speed | Cost | Coverage | Limitation |
|---|---|---|---|---|
| Walking patrol | 2–5 km/day | €30–50/km | Full | Slow, subjective |
| Helicopter | 50–100 km/hr | €200–500/km | Partial | Expensive, weather-dependent |
| Drone (UAV) | 5–15 km/day | €50–150/km | Partial | Needs pilot, regulations |
| Satellite (SAT) | Full coverage | €5–20/km | Limited detail | Low resolution close-up |
| AutoGuard (vehicle) | 30–60 km/hr | €10–30/km | Roadside network | Roadside only |
Most DSOs conduct a full network walk once a year — and even then it is only a visual check. 60–70% of faults remain hidden until the next patrol, or until a failure occurs.
2. The AutoGuard Concept
AutoGuard is a vehicle-mounted camera system that automatically surveys the roadside utility network while driving normally. The concept is simple but powerful:
Utility vehicles drive along the network every day — for fault response, maintenance, and customer service. If these vehicles are equipped with AutoGuard, every trip automatically becomes a condition survey. No extra time, no extra cost — just data.
3. Hardware Setup
AutoGuard uses a dual-camera architecture: one forward-facing and one rear-facing high-resolution camera. This approach captures each pole from multiple angles as the vehicle passes — providing near-3D reconstruction.
Forward camera: Captures the pole as you approach it (front side)
Rear camera: Captures the pole after you pass it (back side)
Result: The moving vehicle records from a continuously changing angle — giving the pole ~350° coverage. The blind spot is just 5–10° on the side furthest from the road.
| Distance | Resolution | Detectable | Example faults |
|---|---|---|---|
| 5–10 m | 5–10 px/cm | Fine details | Cracks, rust spots, missing bolts |
| 10–15 m | 2–5 px/cm | Components | Broken insulator, tilt, deformation |
| 15–25 m | 1–2 px/cm | Structure | Pole type, vegetation proximity |
| 25+ m | <1 px/cm | Presence only | Pole detection, major anomalies |
- LV/MV (0.4–20 kV): 3–8 m → EXCELLENT resolution
- HV (35–110 kV): 8–15 m → GOOD resolution
- EHV (110+ kV): 15–30 m → ACCEPTABLE (these run alongside roads less frequently)
- 80% of roadside poles are within optimal distance for detailed inspection.
4. AI Detection Engine
The heart of AutoGuard is its real-time AI processing. The neural network running inside the vehicle analyses 10–30 frames per second and instantly flags detected objects and faults.
Detection Models
| Model | Architecture | Accuracy | Speed |
|---|---|---|---|
| Pole detector | YOLOv8-L | ~94% mAP | 25 fps |
| Insulator detector | YOLOv8-M | ~94% mAP | 35 fps |
| Tilt estimator | Custom CNN | ±0.5° error | 50 fps |
| Fault classifier | ResNet-50 | ~94% accuracy | 40 fps |
| Vegetation detector | Mask R-CNN | ~94% IoU | 15 fps |
Start With What It Does Well
There is a common mistake in evaluating inspection technology: people start by asking what it cannot do. We think that is the wrong question. Ask first what it does well — then build from there.
5. What Do We Detect?
AutoGuard can identify 15+ object types and fault categories in real time:
Poles
Automatic recognition of wood, concrete, and steel types
98.5%Insulators
String, composite, type and count identification
96.2%Pole Tilt
Deviation from vertical measured in degrees
±0.5°Broken Insulator
Missing or cracked discs
94.1%Bird Nests
On poles and cross-arms
89.5%Corrosion
Rust on steel components
87.2%Conductor Sag
Excessive sag detection
91.8%Pole ID
OCR-based identifier reading
95.6%6. Workflow
The full AutoGuard workflow from route planning to report generation:
7. Comparison: Traditional vs. AutoGuard
✗ Traditional Survey
- 1–2 full walks per year
- Subjective, human assessment
- Paper-based documentation
- No precise GPS position
- Weeks for processing
- €30–50/km cost
- Visible faults only
- No trend analysis
✓ AutoGuard Survey
- Continuous — every service trip
- Objective, AI-based assessment
- Digital, GIS-integrated
- cm-accurate GPS position
- Real-time results
- €10–30/km cost
- Hidden faults too (tilt, vegetation)
- Automated trend analysis
"AutoGuard doesn't replace human expertise — it amplifies it. The system handles the monotonous observation, freeing people to focus on decision-making and action."
For a 5,000 km network, AutoGuard replaces the majority of traditional walking patrols — turning an annual cost of €200,000 into a lean, continuously updated intelligence layer.
€700,000 cumulative savings over 5 years
8. GridGuardian Integration
AutoGuard is not a standalone product — it is an integral part of the GridGuardian platform, working in integrated fashion alongside other monitoring methods (satellite, drone, sensor).
Comprehensive Network Monitoring
AutoGuard is the roadside layer of the GridGuardian ecosystem — complementing all other data sources.
🛰️ SATGuard
Satellite monitoring: large area, vegetation, sag, change detection
🚁 DroneGuard
Drone inspection: detailed imagery, hard-to-reach locations
🚗 AutoGuard
Vehicle-based: roadside, continuous, cost-effective
🧠 ImageGuard
AI processing: unified fault classification from all sources
- Cost-effective: 60–80% savings vs. traditional methods
- Continuous: Every service trip is an inspection
- Objective: AI-based, not subjective human assessment
- Fast: Real-time results, no delay
- Integrated: Part of the full GridGuardian platform
- Scalable: From a single vehicle to an entire fleet
Turn Every Trip Into an Inspection
Request a free AutoGuard demo — we'll show you how it works in practice on your network.
Request a Demo →