Abstract
Abstract. As the majority of the earth population is living in urban environments, cities are continuously evolving and efficient monitoring tools are needed to retrieve and classify their evolution. In this context, analysing changes between two dates is a crucial point. In urban environments, most changes occur along the vertical axis (with new construction or demolition of buildings) and the use of 3D data is therefore mandatory. Among them, LiDAR constitutes a valuable source of information. However, With the difficulty of processing sparse and unordered 3D point clouds, most of existing methods start by rasterizing point clouds (for example to Digital Surface Models) before using more conventional image processing tools. This implies a significant loss of information. Among existing studies dealing directly with point clouds, and to the best of our knowledge, no deep neural network-based method has been explored yet. Thus, in order to fill this gap and to test the ability of deep methods to deal with change detection and characterization of 3D point clouds, we propose a Siamese network with Kernel Point Convolution inspired by Siamese architectures that have already shown their performances on change detection in 2D images and on KPConv network which achieves high-quality results for semantic segmentation of raw 3D point clouds. We show quantitatively and qualitatively that our method outperforms by more than 25% (in terms of average Intersection over Union for classes of change) existing machine learning methods based on hand-crafted features.
Highlights
Due to anthropogenic activities and natural disasters, cities are continuously evolving, yielding critical environmental problems
We proposed a novel deep learning method which takes as input bi-temporal raw Point Clouds (PCs) and gives final results at the 3D point level
Our method is inspired by 2D change detection deep networks using Siamese architecture and deep network used for semantic segmentation in 3D PCs, in particular Kernel Point Convolutions
Summary
Due to anthropogenic activities and natural disasters, cities are continuously evolving, yielding critical environmental problems (e.g. air pollution and heat waves). United Nations report that more than 50% of the earth population is currently living in urban areas Monitoring their evolution is critical and can be achieved with change detection from remote sensing data. The 3D PCs into 3D voxels is somehow better, such a strategy is constrained by the resolution and faces both loss of information and management of sparse voluminous data. This calls for methodologies able to cope with 3D PCs directly, and to distinguish between real changes from those induced by 3D acquisition. As we believe that 3D data could bring much more information to retrieve and classify changes, our goal is to design a deep network able to directly process PCs
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
