By using a LiDAR camera, a point cloud can be generated in a simple manner, which can create a 1-to-1 mapping of an environment or object. With the help of various software packages, specific operations can be performed on these point clouds, such as calculating the distance between points. However, this requires specific knowledge and performing these operations manually takes a significant amount of time. As a result, point clouds are underutilized for calculations, and conventional measurement methods are still relied upon, even though LiDAR technology can generate highly accurate point clouds. By using AI to classify surface types in a point cloud, calculations can also be made very quickly and automatically with these surfaces.

The intended outcome is a framework that can automatically classify and segment surfaces/objects based on training data and, for example, calculate their areas and display the number of them. The goal is to eventually have a self-training AI model through unsupervised learning, which can automatically classify datasets, possibly with the help of a description of the point cloud. To verify and improve classifications and segmentations, a simplistic GUI will also be created. The data that has been verified can then be automatically added to the training data to allow the AI algorithm to learn and become more accurate.

The ultimate outcome of this project will be applicable to a wide range of sectors. For example, there is a high demand in the construction industry to perform measurements of buildings or conduct measurements in large constructions or road works. However, there are also applications in agriculture or the healthcare sector, where point clouds are already being used.

If you are interested in staying informed about this research and/or participating in the user group, please contact us via the form below or through david.vandenbroeck@ucll.be.