The Netherlands has a complex infrastructure given its population of about 17 million people. Average population density is 400 people per square kilometer. What’s more, the majority of the people lives in a relatively small area, namely, the Randstad.

At the same time, the Netherlands are renowned for being one of the most developed countries in the world. Per definition, that means that its underground infrastructure is highly developed. At various places, there are underground cables and pipelines which manage the underground transport of compounds and signals. One could imagine the sewer here, but also telecommunications cables.

Today, the location of cables and pipelines is being administered properly under pressure of existing legislation. In the past, however, this was different, since it was primarily important that infrastructure was laid out. This was especially the case in the period after World War 2 and the 60s of the 20th century.

Excavation damage

In developed countries, the number of construction projects is also high. Construction comes with digging, which has consequences for the underground infrastructure. That is, today there is a substantial amount of excavation damage. In fact, if we make a simple calculation based on the known statistics, a cable or pipeline is hit in our country every 3 minutes. This results in economic damages of tens of millions of Euros per annum. And we then didn’t even discuss the non-economic damages. Think about safety, for example – because what if a gas pipeline is hit?

Thankfully, since about ten years change is happening in the excavation world. After one who intends to dig into the ground hires an excavation company, this company is required to report the activities in the so-called KLIC registry of the Kadaster (i.e., the Dutch land registry). The Kadaster then identifies cables and pipelines are located in the bounded area and solicits with their owners to provide the excavation company with more detailed information about their infrastructure. Additionally, the owner must be physically present during excavation activities when the infrastructure is criticial, e.g. in the case of important telecommunications cables.

However, the KLIC is not complete with regards to older cables and pipelines. Very regularly, orphan cables and orphan pipelines are discovered, as well as wrongly installed utilities. Additionally, there is some tension between the parties responsible for the infrastructure. This means that the odds remain that utilities are struck during excavation work. But what to do about it?

Deployment of geophysical techniques

TerraCarta B.V. from Hoogeveen is a high tech company specializing in mapping the underground infrastructure. By utilizing various measurement techniques and combining those with in-house experts, the company is capable of providing clients with insights regarding what’s in the underground. TerraCarta works for various clients, both from the public and private sector.

Ground penetrating radar (GPR) is one of the measurement techniques utilized by TerraCarta. The device, which emits radar waves into the ground that reflect on underground objects, allows users to map the underground. The so-called GPR images can subsequently be analyzed by TerraCarta experts who use their knowledge to find patterns in the data. This knowledge, which takes years to acquire, can be linked to other insights from physics, which allows them to distinguish various types of cables and pipelines.

Image: a simulated radar image with reflections of underground objects.

TerraCarta + Aime = GPR + Machine Learning

Nevertheless, analyzing GPR imagery is a tough job, even for GPR experts. Machine learning algorithms, which can detect patterns in large amounts of expert-labeled datastes, can amplify expert knowledge by providing some kind of advice with regards to what’s in the underground. It is therefore very interesting to see how TerraCarta’s experts and machine learning models can work together in a symbiotic relationship.

For this reason, Aime started working together with TerraCarta in the beginning of 2019. In this collaboration, especially Christian from Aime works towards obtaining his master’s degree at the University of Twente, by investigating the applicability of state-of-the-art machine laerning algorithms for identifying characteristics of underground cables and pipelines. Secondary, here, is to discover ways of identifying those objects in the first place. We take the role of intelligence amplification, which is a way to deploy machine learning in an advisory role. We therefore do not aim to replace the human, but rathre to amplify their intelligence.

The interest for underground cables and pipelines has emerged from a previous project conducted by Aime, which was unfortunately discontinued by the client. Aime does however believe in her new ideas and is convinced that they can be thoroughly investigated in the current study.

Finishing the literature review

In the past few weeks, we conducted a thorough literature review with which the state-of-the-art in scientific fields related to this topic was identified. It has become clear how intelligence amplification can be put to practice, but more importantly: what has been found and tried over the past few years in the areas of GPR and machine learning. It has become apparent that various steps can still be made.

By working together with TerraCarta experts, we recently created a plan for the subsequent phase. We intend to validate three novel algorithms, with which we study their applicability to the GPR problem. We strive to produce a useful model – or a useful combination of models. We then intend to work towards a Proof-of-Concept analysis tool, which allows TerraCarta experts to obtain more insight based on what has been learnt in the past. Especially the power of machine learning!

Besides TerraCarta, the scientific communities on those topics will also obtain new insights. Do our approaches work properly? Does the tool support the work conducted by the experts? What can be improved? Those are questions to which we will attempt to find answers.

Are you interested in how AI can benefit your organization? Please feel invited to participate in our workshop AI & Big Data in Deventer. Will you visit us as well? 💬