Free GPR Datasets for Underground Utility Detection

Real-world radargrams for locating buried pipes, cables and ducts before excavation

Locating buried utilities before excavation is one of the most safety-critical applications of Ground Penetrating Radar. Every year, thousands of utility strikes cause service interruptions, injuries, and costly repairs. GPR provides a non-destructive way to detect, map, and characterize underground pipes, cables, and ducts before any ground disruption.

GPRbase provides free datasets acquired on real utility surveys: urban streets, industrial sites, road crossings, and greenfield investigations. Each dataset contains raw radargrams in GSSI format (.DZT), documented with acquisition context and target information. Data is distributed under CC BY-NC-SA 4.0 for free use in training, research, software development, and public awareness programs.

These datasets support utility detection training, SUE (Subsurface Utility Engineering) skill development, machine learning research on utility classification, and the validation of automated detection algorithms. Real subsurface conditions are represented: multi-utility corridors, deep drainage, mixed materials (metal, PVC, concrete), and challenging soil types.

GPR for underground utility detection

GPR detects both conductive utilities (metal pipes, cables) and non-conductive utilities (PVC, concrete, clay) via reflections at material interfaces. Its advantage over locators lies in detecting non-metallic pipes and characterizing depth. It complements electromagnetic locators, providing a full picture of the subsurface for damage prevention and mapping campaigns.

Metallic pipe and cable detection

Metal utilities produce strong, easily identifiable hyperbolic signatures on radargrams. Modern GPR software automatically picks these hyperbolas to measure cover depth, spacing, and orientation. This information supports damage prevention protocols and helps map complex utility corridors before infrastructure work.

Non-metallic utility detection

PVC water pipes, concrete drainage lines, and clay conduits produce weaker but still distinctive reflections. Detection depends on the dielectric contrast with surrounding soil — a dry soil around a water-filled PVC pipe gives strong contrast, while saturated soil may attenuate the signal significantly. Multi-frequency surveys often help resolve difficult cases.

Depth and mapping accuracy

GPR depth accuracy is typically ±10% with proper velocity calibration through known-depth targets. Horizontal positioning accuracy reaches centimeter-level precision with cart-mounted systems using odometers or GPS. This makes GPR suitable for documented as-built utility maps required by regulatory frameworks.

GPR frequencies for utility detection

Medium frequencies from 200 to 400 MHz are the most suitable for typical utility depths of 2 to 5 meters. Higher frequencies (600-900 MHz) improve resolution for shallow utilities and dense corridors. Lower frequencies (100 MHz) may be used for very deep infrastructure but sacrifice resolution.

Download raw utility GPR data

All datasets are raw .DZT files with no processing overlay. This makes them ideal for utility interpretation training, ML model development for automated utility classification, or software validation. Free download, immediate access, no account creation required.

Frequently asked questions — Utility GPR

Can GPR find PVC pipes?

Yes, when there is dielectric contrast between the pipe and surrounding soil. Deeper or wet-soil situations may reduce detection performance. Water-filled PVC pipes are typically easier to detect than empty ones.

What is the maximum depth for utility detection?

Depends on soil and frequency. Typical performance: 3-5 m in favorable conditions with a 400 MHz antenna; up to 8-10 m with 200 MHz in dry sandy soil. Wet clay strongly limits depth.

Do these datasets include GPS positions?

When available, .DZG files with GPS traces are included alongside the .DZT radargrams. The dataset metadata specifies whether GPS is present.

Can I use these for autonomous vehicle or robot sensor testing?

Yes. The raw signal data is well-suited for sensor fusion research and autonomous perception development, under CC BY-NC-SA 4.0.

Are the datasets georeferenced?

Some are, with precise coordinates when the contributor permits. Others use generic location tags (e.g., "France") to protect site confidentiality. Metadata indicates the level of georeferencing for each dataset.

Have utility survey data to share?

Every contributed dataset helps improve utility detection training and research worldwide. Contributions from surveyors, utility mappers, and researchers enrich the community.

Submit your data