Abstract:3D point clouds of natural environments relevant to problems in geomorphology (rivers, coastal environments, cliffs,…) often require classification of the data into elementary relevant classes. A typical example is the separation of riparian vegetation from ground in fluvial environments, the distinction between fresh surfaces and rockfall in cliff environments, or more generally the classification of surfaces according to their morphology (e.g. the presence of bedforms or by grain size). Natural surfaces are heterogeneous and their distinctive properties are seldom defined at a unique scale, prompting the use of multi-scale criteria to achieve a high degree of classification success. We have thus defined a multi-scale measure of the point cloud dimensionality around each point. The dimensionality characterizes the local 3D organization of the point cloud within spheres centered on the measured points and varies from being 1D (points set along a line), 2D (points forming a plane) to the full 3D volume. By varying the diameter of the sphere, we can thus monitor how the local cloud geometry behaves across scales. We present the technique and illustrate its efficiency in separating riparian vegetation from ground and classifying a mountain stream as vegetation, rock, gravel or water surface. In these two cases, separating the vegetation from ground or other classes achieve accuracy larger than 98%. Comparison with a single scale approach shows the superiority of the multi-scale analysis in enhancing class separability and spatial resolution of the classification. Scenes between 10 and one hundred million points can be classified on a common laptop in a reasonable time. The technique is robust to missing data, shadow zones and changes in point density within the scene. The classification is fast and accurate and can account for some degree of intra-class morphological variability such as different vegetation types. A probabilistic confidence in the classification result is given at each point, allowing the user to remove the points for which the classification is uncertain. The process can be both fully automated (minimal user input once, all scenes treated in large computation batches), but also fully customized by the user including a graphical definition of the classifiers if so desired. Working classifiers can be exchanged between users independently of the instrument used to acquire the data avoiding the need to go through full training of the classifier. Although developed for fully 3D data, the method can be readily applied to 2.5D airborne lidar data.

A Method of Point Cloud Classification Using Multi-scale Dimensionality Features and Transductive Learning

Classification of LiDAR Point Clouds Using Supervoxel-Based Detrended Feature and Perception-Weighted Graphical Model.

A Multilevel Point-Cluster-Based Discriminative Feature for Als Point Cloud Classification

Discriminative-Dictionary-Learning-Based Multilevel Point-Cluster Features for ALS Point-Cloud Classification

A Multiscale and Hierarchical Feature Extraction Method for Terrestrial Laser Scanning Point Cloud Classification

A Three-Step Approach for TLS Point Cloud Classification.

Classification of Urban Point Clouds: A Robust Supervised Approach with Automatically Generating Training Data

Using Multi-Scale and Hierarchical Deep Convolutional Features for 3D Semantic Classification of TLS Point Clouds.

Point Set Multi-Level Aggregation Feature Extraction Based on Multi-Scale Max Pooling and LDA for Point Cloud Classification

Three-Dimensional Point Cloud Classification of Large Outdoor Scenes Based on Vertical Slice Sampling and Centroid Distance Histogram

Classifying airborne LiDAR point clouds via deep features learned by a multi-scale convolutional neural network

3D terrestrial lidar data classification of complex natural scenes using a multi-scale dimensionality criterion: Applications in geomorphology

Deep point embedding for urban classification using ALS point clouds: A new perspective from local to global

Segmentation-based Classification for 3D Point Clouds in the Road Environment

A Hierarchical Approach for Point Cloud Classification with 3D Contextual Features.

Intelligent Classification of Point Clouds for Indoor Components Based on Dimensionality Reduction

Tree Species Classification of Point Clouds from Different Laser Sensors Using the PointNet++ Deep Learning Method

ALS Point Cloud Classification with Small Training Data Set Based on Transfer Learning

Classify Tree Species from Point Clouds Generated by Different Laser Sensors: A Multi-View Projection Strategy

ALS Point Cloud Classification by Integrating an Improved Fully Convolutional Network into Transfer Learning with Multi-Scale and Multi-View Deep Features