IMQS has developed an Automated Road Assessment solution to overcome road surface assessment challenges. Read more to learn how to overcome issues of cost, time and data accuracy in road surface surveying.


The development and maintenance of road networks play an important role in the economic development of a country. The proper upkeep of roads improves their safety, reduces the cost of transportation, both in terms of money and time, and enables better regional and international economic integration.

South Africa maintains a total road network of approximately 535 000 km. Of this, 366 872 km are non-urban and 168 000 km are urban.

Road-surface assessment surveys rely on teams of people to manually assess road surfaces according to a list of criteria. Teams walk down every road, taking note of the road surface quality, and documenting various markers, such as cracks and potholes, in the process of mapping the road network’s quality. Common criteria include: the adherence, the micro-texture, the macro-texture, and the surface degradation of roads. The resulting information is integral in the better planning of road development and maintenance.

Presently, the financial and temporal costs of manual road-surface assessments can be staggering. It can take a human up to an hour to cover just one kilometre. In South Africa, this would amount to 168 000 hours of labour in urban areas alone. The manual approach is highly subjective, qualitative, and sometimes inaccurate. Evaluation results may vary due to personal judgment, distress type, or severity. The subjective nature of this work may, moreover, impact the data, rendering it unreliable and misleading. Unreliable data can result in non-optimal allocation of agency funding and resources. It is therefore imperative for agencies to ensure that high quality road condition data is collected and processed.

Currently, available technological solutions that cater for condition assessment are expensive, especially in developing markets where budgets are limited. These solutions often consist of expensive equipment and specialised vehicles. Special skills are required to operate the technological solutions. Finally, a large time lapse exists between condition assessment and acquiring the resulting data.


Reliable and cost-effective automation can help overcome the above challenges. IMQS Software has developed its own automated road-assessment solution by harnessing the power of machine learning and image recognition. This solution requires no specialised skills, and considerably increases the speed of acquiring reliable road-condition data.

Machine learning is an application of artificial intelligence (AI) that provides systems the ability to automatically learn and improve from experience without being explicitly programmed. Computer programs are developed that can access data and use it to learn for themselves. The process of learning begins with observations or data, such as examples, direct experience, or instruction, in order to look for patterns in data and make better decisions in the future based on the examples that are provided. The primary aim is to allow the computers to learn automatically without human intervention or assistance and adjust actions accordingly.

IMQS has successfully merged machine learning with image recognition to offer municipalities a solution that substantially reduces the time and cost of performing road-surface surveys.


IMQS’s Automated Road Surface Assessment process consists of three integrated horizontals:

Horizontal 1: Road Survey

The first part of the process is data collection. Data includes video footage of roadd surfaces and their corresponding GPS coordinates. A Fuji XD2 DSLR camera is mounted to tripod fastened to a vehicle via suction cups. The camera is pointed to the road, capturing road data according to preconfigured parameters / criteria. The driver of the vehicle activates the video stream and can capture an estimated average of 30 km of road surface per hour. Due to the lack of integrated GPS in high quality DSLR cameras, IMQS developed an application to simultaneously capture GPS information during the job. The application and camera are simultaneously activated and information is merged in the cloud in order to geographically establish the GPS of road segments and the faults associated with them.

Horizontal 2: Cloud-based data processing

The resulting video footage and GPS data are uploaded to processed in the cloud.

The video footage is deconstructed into single frames, which are then flattened and stitched together to build a flat image of the road surface. Each reconstructed representation of the road surface offers a 2mm per pixel resolution for accurate and in-depth analysis. The GPS data is lined to the video data to build a concise geographic representation of road-surface segments that can be analysed by an asset manager.

IMAGE 1: Original Image

IMAGE 2: Angled and flattened

IMAGE 3: 50 images stitched together

A Convolutional Neural Network (CNN) is trained to find relevant features in the road surface, such as cracks and potholes, according to South Africa’s TMH-9 Visual Assessment of Roads. CNNs have proven to be useful in analysing visual imagery. They make use of sequential layers and an input file that contains the “weights” of the parameters. Compared to Fully Connected Networks, CNN’s rely on fewer parameters and are cheaper in terms of memory and computing power. When harnessing the power of CNN, IMQS starts from a clean slate and trains the network to understand what it sees in the camera footage. As the machine learns, the process becomes more reliable and the data ever more accurate.

In the context of road assessment, training involves teaching an artificial intelligence to recognise and mark road-surface incidents according to TMH-9 criteria. Once the flat earth model of a municipality’s road surface has been produced, a trained individual will label the block where an item occurs – incident blocks. The model is then used to train the machine to locate such incidents independently.

Horizontal 3: Asset Management

The captured and assessed information is finally used to build an accurate map of the road network, which shows, with precision, the surface quality on every part of the road. The information is geographically presented in IMQS and can be used by infrastructure asset managers to gain a complete and verified understanding of their road network’s condition.


While no specialised skills are needed to operate the hardware, challenges exist when implementing such internet-driven solutions in developing economy municipalities.

The first challenge is the requirement for high-speed internet to upload large amounts of data to the cloud and process it - about 300 GB per day. This is especially a problem for developing market economies and rural communities. A second challenge is that of surveying gravel road surfaces. Not only do gravel roads have their own assessment criteria, but are currently a challenge when it comes to mounting hardware to the surveying vehicle.

On the other hand, the benefits far outweigh the challenges.

Firstly, the IMQS Automated Road Surface Assessment solution could decrease cost by an estimated factor of ten, while increasing survey speed to roughly 30 km an hour. Secondly, the resulting accuracy of data also means that more effective and timely decisions can be made regarding maintenance and management of municipal road segments and networks. Finally, there positive spinoffs include the simultaneous mapping of road segments during data acquisition, as well as the potential to survey road furniture with similar methods.

This IMQS solution is built to seamlessly integrate with existing geographically enabled IMQS products, such as the IMQS Asset Register and Project Control System, but will also integrate with a municipality’s preferred asset management system. In so doing, it buttresses the existing integrated lifecycle asset management approach that makes IMQS a leader in the infrastructure asset management industry.