Home Home Pavement Distress Identification

Pavement Distress Identification

An efficient and effective solution to identify pavement distresses for pavement maintenance and management decisions.

Pavement Management Systems (PMS) require reliable pavement condition data to conduct needs analysis for appropriate maintenance and rehabilitation treatments. To address the data collection safety, efficiency, and subjectivity concerns, roadway agencies have tried to move away from manual windshield and walking surveys to more automated data collection technologies.


With the advent of advanced sensor technologies and progress in machine vision, the industry has adopted a semi-automated approach. Several pavement distress types such as cracking and rutting are automatically detected using machine learning techniques and numerical computations on pavement images and sensor data. However, certain other pavement distress types are still manually identified based on automatically collected images. Despite significant advancements in relevant research, the current state of the practice is still semi-automated regarding detection of potholes, patching, raveling, weathering, and alike.

State of the Art

This solution demonstrates integration of the state-of-the-art artificial intelligence technologies for automated real-time identification of pavement distresses like potholes and patching from forward-facing roadway images. This ongoing project uses one of the very latest machine learning algorithms to achieve superior effectiveness, efficiency, repeatability and reproducibility. Regardless of the image or video capturing device, the media files can be uploaded to the cloud and processed quickly to generate a geo-located map of pothole locations on the road network.

Supported Interfaces


For transportation agencies that use multi-purpose survey vehicles for data collection, our solution packaged in an edge device can be utilized to process raw roadway image data in real-time. By utilizing the edge device solution, analyzing data closer to where the data is collected, provides huge benefits to transportation agencies in terms of cost and time of uploading data to the cloud.


The solution is available as a cloud computing web-service. The web-service provides a user interface as well as APIs where raw roadway images can be uploaded and processed quickly using machine learning techniques. The processed results are available in Comma Separated Values (CSV) and Keyhole Markup Language (KMZ) formats. Agencies can input the downloaded KMZ files to their existing GIS software to view the identified distresses on a map.


The solution can be deployed in a smartphone application for real-time processing of the roadway images. The application can use the smartphone’s camera to collect raw images and process the images using machine learning models. The results can be uploaded to the cloud in real-time as soon as they are processed using mobile data or they can be uploaded in bulk using a WIFI connection when it is available.