Ville Lehtola

Dr.Sc., Assistant Professor, University of Twente, the Netherlands

Autonomous ships, perception

Indoor 3D, SLAM, and reconstruction

Robotics, perception, SLAM

LIDAR signal processing, spatial correlations

My research field is in (machine) perception. I work with sensors and sensor data. My two main research interests are (1) the autonomy of systems and (2) measurements done from motion.

My dream is to have the world represented in a digital reality.

Some more or less randomly selected works below:

LIDAR sensors can yield massive data streams. These data can be reduced to make all following processing easier, by pre-classifying noise and solid (=ground/buildings/tree trunks) and non-solid surfaces (=vegetation), with spatial correlations. These techniques remedy the problem with noisy LIDARs, e.g. single photon LIDARs, which are vulnerable to background illumination (=sunlight).

Lehtola, Ville V., et al. "Preregistration classification of mobile LIDAR data using spatial correlations." IEEE transactions on geoscience and remote sensing 57.9 (2019): 6900-6915.

Laser scanning indoor 3D point clouds and capturing imagery are important for reconstructing Building Information Models (BIM). This is a highly cited paper:

Lehtola, Ville V., et al. "Comparison of the selected state-of-the-art 3D indoor scanning and point cloud generation methods." Remote sensing 9.8 (2017): 796.

Autonomous ships are soon less scifi and more about reality. AI plays a role. We review techniques how ships can be given super-human eyes and ears.

Thombre, S., ...., Lehtola, V. V. (2020). Sensors and AI Techniques for Situational Awareness in Autonomous Ships: A Review. IEEE Transactions on Intelligent Transportation Systems, (Early access)

https://doi.org/10.1109/TITS.2020.3023957

E-navigation is essential for autonomous ships. Route planning (here: A* graph search) needs to account for multi-modal data and occurs before the ship leaves port.

Lehtola, V., Montewka, J., Goerlandt, F., Guinness, R., & Lensu, M. (2019). Finding safe and efficient shipping routes in ice-covered waters: A framework and a model. Cold regions science and technology, 165, 102795.

Autonomy of forest machines requires perception. Here, we use a line scanner and joystick controls to estimate the posture of the crane of a prototype forest machine. Simultaneously, a 3D point cloud of the environment is acquired.

Hyyti, H., Lehtola, V. V., & Visala, A. (2018). Forestry crane posture estimation with a two‐dimensional laser scanner. Journal of Field Robotics, 35(7), 1025-1049.

Umbra3D Ltd (a NVIDIA spin-off company) creates digital scenes for games and movie industry. We studied the computer vision problem of unocclusion for 3D pointclouds.

https://blog.umbra3d.com/blog/fixing-laser-scans-with-deep-learning

Väänänen, P., & Lehtola, V. (2019). Inpainting occlusion holes in 3d built environment point clouds. The International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, 42, 393-398.