What is a PCI Proxy?

PCI Proxy, Visual Condition Grade Approximation

Definition and Need

A PCI proxy (visual condition grade approximation) is a condition estimate derived from visible image features of the pavement surface, used in place of the Pavement Condition Index (PCI) when a full distress survey is not available. In research and in some agency pilots it is usually reported as a coarse ordinal grade, for example from 1 (good) to 5 (serious), rather than the 0 to 100 PCI. It approximates the index without requiring a field distress survey, deduct value computation, or the sampling protocol of ASTM D5340 (airports) or ASTM D6433 (roads and parking lots).

TarmacView has not built a PCI proxy.

The need for proxies comes from a tension in pavement management. The official PCI survey gives the most authoritative condition data, but it is labor-intensive and slow. The standard requires trained inspectors to walk each sample unit, identify the distress types from the standard list, assign low, medium, or high severity to each, measure quantities, convert them to densities, and apply deduct value curves. A survey covering the runways, taxiways, and aprons of one airport can take weeks of fieldwork, and a large road network may be surveyed in full only once every few years.

A proxy addresses the gap between those full surveys. Because it works from images that can be collected by drive-over vehicles or drones, it can be refreshed more often and cover more pavement. It is a screening tool that points the full survey at the sections that most need it, not a replacement for the survey.

PCI Proxy vs Official PCI

The official PCI is a cardinal rating from 0 to 100 that quantifies the structural integrity and operational surface condition of a pavement. The method was developed by the U.S. Army Corps of Engineers in the 1970s for airfield pavement management and later adopted by ASTM International for roads, parking lots, and, through ASTM D5340, airports.

The official process divides the network into branches, sections, and sample units. For each sample unit the inspector records every distress present, such as fatigue (alligator) cracking, block cracking, longitudinal and transverse cracking, rutting, raveling, patching, and potholes, with a severity of low, medium, or high and a measured quantity. Density is the quantity divided by the sample unit area. Deduct values are read from standardized curves for each distress, severity, and density, then adjusted by a correction procedure that accounts for several distresses acting together. The PCI is 100 minus the maximum corrected deduct value, and it maps to a rating from Good down to Failed.

A proxy departs from this in three ways. First, many proxies use an ordinal scale (a few grades) rather than a cardinal 0 to 100 scale. A grade 3 pavement is worse than grade 2, but the steps are not equal in any measurable unit, which is acceptable when the aim is ranking sections instead of precision. Second, a proxy derives its grade from image-derived features instead of field measurement of every distress type and severity. Third, a proxy may use rule tables or a trained model instead of deduct value curves, which changes how easily the mapping can be audited.

How Proxies Are Built

Most proxies start from features that can be read from a surface image. The common ones are the share of the surface covered by cracks, the crack type (alligator, block, longitudinal, transverse), the crack width, and the presence of patching, potholes, raveling, or visible deformation. Some proxies add a holistic surface quality score that captures texture, aggregate condition, and oxidation color, which crack counts alone miss.

Two construction styles are common. Rule-based proxies map the features to a grade with a published table, so a reviewer can see why a section received its grade and recalibrate the thresholds for local materials and traffic. Learned proxies train a model on images graded by experts. A frequently used research approach is to compute an embedding of each image with a self-supervised vision model, then assign the grade by comparing it with the nearest graded reference images, a k-nearest-neighbor vote. Learned proxies can capture surface texture that rules miss, but they are harder to audit, and they depend on the training images matching the pavement type and imaging conditions in use.

Crack area and width usually come from pixel-level segmentation of the cracks, followed by skeletonization and distance measurements across the crack to estimate width. Reported accuracy for width estimation in controlled studies is around 90 percent (Ibragimov et al., 2024, Sensors), and it drops with blur, shadows, and surface texture.

How Proxies Are Validated Against ASTM D5340 PCI

A proxy is only as good as its agreement with the official index. Validation pairs each proxy grade with the PCI measured on the same sample unit, so that images and field surveys cover the same pavement. For airports, D5340 sample units are about 20 slabs for concrete or about 5,000 square feet for asphalt, which fixes the unit of comparison.

Typical validation measures are rank correlation between proxy grade and PCI, the share of sample units where the proxy lands within one grade of the PCI rating band, and a confusion matrix that shows where the proxy confuses neighboring grades. Because the two are measured at different times, the gap between image capture and survey matters, and so does a check for distresses the proxy cannot see. A proxy validated on one network, climate, and pavement type does not carry over automatically to another, so local calibration against a sample of full surveys is the usual practice.

Crack Width Bands Used in Proxies

Proxies that use crack width commonly follow bands derived from AASHTO PP44 and PP67, which set protocols for quantifying cracks in asphalt pavement surfaces. The bands are consistent with the severity classes of the FHWA Long-Term Pavement Performance (LTPP) Distress Identification Manual.

Width addresses a known weakness of crack area percentage alone. Two pavements with the same 10 percent crack area can be in very different condition if one has only hairline thermal cracks and the other has wide alligator cracking. Giving wider cracks more weight is a common way to separate the two cases.

Limitations, No Binder Content, No Density, No Water Damage Progression

A PCI proxy has fundamental limitations. They are not flaws of a particular product. They are the boundary of what surface images can show.

Binder content is the proportion of asphalt binder in the mix. It is determined in the laboratory by extraction (AASHTO T164, ASTM D2172) or ignition (AASHTO T308, ASTM D6307) tests. Visual inspection cannot tell an optimum mix from a binder-rich one that may bleed or rut, or a binder-lean one that may ravel and crack. Surface discoloration suggests oxidation but does not quantify the remaining binder properties.

Pavement density, the in-place density or air voids of the asphalt layer, cannot be measured from surface images. Air voids below about 3 percent risk bleeding and rutting, while air voids above about 8 percent risk accelerated oxidation, moisture damage, and cracking. Density comes from nuclear density gauges (ASTM D2950), cores tested in the laboratory (ASTM D2726, D6752), or non-destructive electromagnetic methods.

Water damage progression, the stripping of binder from aggregate, starts inside the pavement and is invisible at the surface until it has progressed to raveling, potholing, or cracking. Moisture susceptibility is evaluated with AASHTO T283 or Hamburg wheel track testing, and stripping in an existing pavement is confirmed by coring.

Additional limitations include structural capacity (FWD or HWD deflection testing), subgrade condition, load transfer efficiency at joints, layer thickness (ground penetrating radar, coring, or construction records), and remaining service life (mechanistic-empirical analysis). A proxy reflects only visible surface condition. It is a functional indicator, not a structural evaluation tool.

Communicating a Proxy Honestly

A proxy is easier to trust when it explains itself. Good practice is to report the input values behind each grade, the rule or model version that produced it, and a confidence indicator that flags poor image quality, bad lighting, or ambiguous features for human review instead of reporting a definitive grade. Transparency lets engineers validate the proxy against their own judgment, recalibrate thresholds for local conditions, and audit unexpected grades.

Every proxy result should carry the same caveats as the limitations above. The grade is derived from visible surface features only, binder content, density, water damage, and structural capacity are not measured, and the proxy should not substitute for laboratory or deflection testing when those properties matter. It is a screening tool, and grades should be validated by field inspection before they drive final engineering decisions, budget allocation, or regulatory compliance.

When a Full PCI Survey is Needed

A proxy suits rapid screening, frequent monitoring, and network-level prioritization. A full ASTM D5340 or D6433 survey is needed in specific circumstances.

Regulatory compliance is the most common reason. For airport pavements in the United States, FAA Advisory Circular 150/5380-7B (Airport Pavement Management Program) calls for PCI surveys of runways, taxiways, and aprons at federally obligated airports, following ASTM D5340 with qualified inspectors. A proxy cannot substitute for that requirement, though it can add condition snapshots between official surveys. U.S. Army Corps of Engineers pavement management for military airfields similarly relies on full PCI surveys at set intervals.

Major capital investment decisions such as rehabilitation, reconstruction, or extension warrant a full survey, because treatment selection and quantity estimation need distress-specific data. The deduct value analysis shows which distresses drive the score and so which ones the chosen treatment must address.

Forensic investigation of premature failure needs distress type, severity, and density data, combined with coring, laboratory testing, and structural testing. A proxy alone cannot support it.

Pavement and overlay design needs structural capacity data that neither the PCI nor a proxy provides, and some design procedures take the cardinal PCI score as an input, which an ordinal proxy cannot supply.

Disagreement between proxy assessments is a trigger too. When several images of the same section produce grades that differ by more than two ordinal grades, either the section is highly variable or image quality is distorting the result, and a full survey settles it.

Acceptance testing of new construction relies on standardized tests for density, smoothness, thickness, and strength, not on condition grades.

The proxy and the full survey are complementary tools. A pavement management program that screens often with a proxy and surveys in depth where it matters gets the best balance of data quality, coverage, and cost.