Defence of doctoral thesis in the field of Computer Science, DI Matti Lehtomäki

Title of the thesis is: "Detection and Recognition of Objects from Mobile Laser Scanning Point Clouds: Case Studies in a Road Environment and Power Line Corridor"
CS_defence_3 photo by Matti Ahlgren

How to automate the interpretation of point cloud data collected using a mobile laser scanner?

How to map, inventory, maintain and plan the road environment or power line corridor as accurately as possible? How to create accurate and detailed maps for the needs of self-driving cars? How to create virtual three-dimensional (3D) city models? And how could all this be done cost-effectively?

Mobile laser scanning and the 3D point clouds it produces can help solve these issues. The point cloud contains point measurements from surfaces, such as building façades and traffic signs. For example, a laser scanner mounted on the roof of a car produces point cloud data with a local accuracy in the millimetre range at best and a global accuracy in the centimetre range. The data is very dense and can contain up to thousands of measurements per square metre. Hundreds of kilometres of road can be mapped daily.

Dense measurements mean large amounts of data – processing of which is laborious. The doctoral thesis seeks to identify whether the interpretation of point clouds can be automated. Case studies were conducted in a road environment and power line corridor.

The dissertation studies the detection of polelike objects such as traffic signs and lampposts from 3D point cloud data in a road environment. The published results are among the first, and they suggest at least slightly higher detection accuracies than previous studies. The thesis also studies new types of object recognition methods, and by applying so-called local descriptor histograms, object classification accuracy was increased by approximately 10 percentage points. The dissertation also demonstrates how mobile laser scanning can be applied to the mapping of power lines outside the road network. The results suggest that the accuracy of the automated power line mapping is higher than 93%.

The results of the dissertation can help practitioners find new, more accurate and efficient solutions for their applications. The dissertation's error analyses help to understand the limitations of the mobile laser scanning technology, and to open new directions for development and research. The dissertation contains promising results on the accuracy of the automated interpretation of the point cloud data.

Opponent: professor François Goulette, Mines Paristech, France

Custos: professor Jouko Lampinen, Aalto University School of Science, Department of Computer Science

Contact information of the doctoral student: [email protected]

The defence will be organised via remote technology (Zoom). Link to the event

Zoom Quick Guide

The doctoral thesis will be publicly displayed 10 days before the defence in the Aaltodoc publication archive of Aalto University.

Electronic thesis. (

  • Published:
  • Updated: