Tree segmentation and change detection of large urban areas based on airborne LiDAR

As the utilization of LiDAR (Light Detection and Ranging) is getting more affordable and available for a wider audience, the analysis of point clouds constructed by laser scanning is earning more attention. Airborne LiDAR is especially useful in the analysis and classification of land objects. We are able to determine if they are natural or artificial objects and what changes occurred to them throughout time by examining multi-temporal data. The goal of our research was to define a completely automatized methodology for the segmentation of vegetation (specifically trees) in urban environment, followed by the qualification and quantification of change detection. Our proposed algorithm provides a robust approach designed to scale dynamically to large areas, in contrast to existing methods that require manual or semi-supervised human interaction and can only be applied on relatively small areas. The algorithm was tested on parts of the Dutch and the Estonian altimetry archives, point cloud datasets that provide several terabytes of data. It was proved to be an effective method for the qualified and quantified change detection of trees, including height and volume changes.

Automated summary

With the increasing affordability and accessibility of LiDAR technology, the analysis of point clouds constructed by laser scanning is becoming more prominent. Airborne LiDAR is particularly valuable for land object analysis and classification, while our research aims to define an automated methodology for vegetation segmentation and change detection in urban areas. The algorithm proposed in our study proved to be effective for qualified and quantified change detection of trees, including height and volume changes, providing a robust approach that can scale dynamically to large areas.