Full-waveform LiDAR Echo Decomposition Based on Dense and Residual Neural Networks

2021 ◽  
Author(s):  
Gangping Liu ◽  
Jun Ke
Keyword(s):  
Neural Networks ◽  
Full Waveform ◽  
Waveform Lidar ◽  
Full Waveform Lidar

scholarly journals A Comparative Study about Data Structures Used for Efficient Management of Voxelised Full-Waveform Airborne LiDAR Data during 3D Polygonal Model Creation

2021 ◽  
Vol 13 (4) ◽  
pp. 559
Author(s):  
Milto Miltiadou ◽  
Neill D. F. Campbell ◽  
Darren Cosker ◽  
Michael G. Grant
Keyword(s):  
Data Structure ◽  
Data Structures ◽  
Airborne Lidar ◽  
Memory Allocation ◽  
Lidar Data ◽  
Full Waveform ◽  
Waveform Lidar ◽  
Airborne Lidar Data ◽  
Full Waveform Lidar ◽  
Polygonal Model

In this paper, we investigate the performance of six data structures for managing voxelised full-waveform airborne LiDAR data during 3D polygonal model creation. While full-waveform LiDAR data has been available for over a decade, extraction of peak points is the most widely used approach of interpreting them. The increased information stored within the waveform data makes interpretation and handling difficult. It is, therefore, important to research which data structures are more appropriate for storing and interpreting the data. In this paper, we investigate the performance of six data structures while voxelising and interpreting full-waveform LiDAR data for 3D polygonal model creation. The data structures are tested in terms of time efficiency and memory consumption during run-time and are the following: (1) 1D-Array that guarantees coherent memory allocation, (2) Voxel Hashing, which uses a hash table for storing the intensity values (3) Octree (4) Integral Volumes that allows finding the sum of any cuboid area in constant time, (5) Octree Max/Min, which is an upgraded octree and (6) Integral Octree, which is proposed here and it is an attempt to combine the benefits of octrees and Integral Volumes. In this paper, it is shown that Integral Volumes is the more time efficient data structure but it requires the most memory allocation. Furthermore, 1D-Array and Integral Volumes require the allocation of coherent space in memory including the empty voxels, while Voxel Hashing and the octree related data structures do not require to allocate memory for empty voxels. These data structures, therefore, and as shown in the test conducted, allocate less memory. To sum up, there is a need to investigate how the LiDAR data are stored in memory. Each tested data structure has different benefits and downsides; therefore, each application should be examined individually.

scholarly journals Complexity and Dynamics of Semi-Arid Vegetation Structure, Function and Diversity Across Spatial Scales from Full Waveform Lidar

2020 ◽  
Author(s):  
Ginikanda Yapa Mudiyanselage Nayani Thanuja Ilangakoon
Keyword(s):  
Structure Function ◽  
Functional Diversity ◽  
Vegetation Structure ◽  
Structural Features ◽  
Airborne Lidar ◽  
Arid Ecosystems ◽  
Full Waveform ◽  
Waveform Lidar ◽  
Full Waveform Lidar ◽  
Semi Arid

Semi-arid ecosystems cover approximately 40% of the earth's terrestrial landscape and show high dynamicity in ecosystem structure and function. These ecosystems play a critical role in global carbon dynamics, productivity, and habitat quality. Semi-arid ecosystems experience a high degree of disturbance that can severely alter ecosystem services and processes. Understanding the structure-function relationships across spatial extents are critical in order to assess their demography, response to disturbance, and for conservation management. In this research, using state-of-the-art full waveform lidar (airborne and spaceborne) and field observations, I developed a framework to assess the complexity and dynamics of vegetation structure, function and diversity across spatial scales in a semi-arid ecosystem. Difficulty in differentiating low stature vegetation from bare ground is the key remote sensing challenge in semi-arid ecosystems. In this study, I developed a workflow to differentiate key plant functional types (PFTs) using both structural and biophysical variables derived from the full waveform lidar and an ensemble random forest technique. The results revealed that waveform lidar pulse width can clearly distinguish shrubs from bare ground. The models showed PFT classification accuracy of 0.81-0.86% and 0.60-0.70% at 10 m and 1 m spatial resolutions, respectively. I found that structural variables were more important than the biophysical variables to differentiate the PFTs in this study area. The study further revealed an overlap between the structural features of different PFTs (e.g. shrubs from trees). Using structural features, I derived three main functional traits (canopy height, plant area index and foliage height diversity) of shrubs and trees that describe canopy architecture and light use efficiency of the ecosystem. I evaluated the trends and patterns of functional diversity and their relationship with non-climatic abiotic factors and fire disturbance. In addition to the fine resolution airborne lidar, I used simulated large footprint spaceborne lidar representing the newly launched Global Ecosystem Dynamics Investigation system (GEDI, a lidar sensor on the International Space Station) to evaluate the potential of capturing functional diversity trends of semi-arid ecosystems at global scales. The consistency of diversity trends between the airborne lidar and GEDI confirmed GEDI's potential to capture functional diversity. I found that the functional diversity in this ecosystem is mainly governed by the local elevation gradient, soil type, and slope. All three functional diversity indices (functional richness, functional evenness and functional divergence) showed a diversity breakpoint near elevations of 1500 m - 1700 m. Functional diversity of fire-disturbed areas revealed that the fires in our study area resulted in a more even and less divergent ecosystem state. Finally, I quantified aboveground biomass using the structural features derived from both the airborne lidar and GEDI data. Regional estimates of biomass can indicate whether an ecosystem is a net carbon sink or source as well as the ecosystem's health (e.g. biodiversity). Further, the potential of large footprint lidar data to estimate biomass in semi-arid ecosystems are not yet fully explored due to the inherent overlapping vegetation responses in the ground signals that can be affected by the ground slope. With a correction to the slope effect, I found that large footprint lidar can explain 42% of variance of biomass with a RMSE of 351 kg/ha (16% RMSE). The model estimated 82% of the study area with less than 50% uncertainty in biomass estimates. The cultivated areas and the areas with high functional richness showed the highest uncertainties. Overall, this dissertation establishes a novel framework to assess the complexity and dynamics of vegetation structure and function of a semi-arid ecosystem from space. This work enhances our understanding of the present state of an ecosystem and provides a foundation for using full waveform lidar to understand the impact of these changes to ecosystem productivity, biodiversity and habitat quality in the coming decades. The methods and algorithms in this dissertation can be directly applied to similar ecosystems with relevant corrections for the appropriate sensor. In addition, this study provides insights to related NASA missions such as ICESat-2 and future NASA missions such as NISAR for deriving vegetation structure and dynamics related to disturbance.

scholarly journals Terrain surfaces and 3-D landcover classification from small footprint full-waveform lidar data: application to badlands

2009 ◽  
Vol 6 (1) ◽  
pp. 151-205 ◽  
Author(s):  
F. Bretar ◽  
A. Chauve ◽  
J.-S. Bailly ◽  
C. Mallet ◽  
A. Jacome
Keyword(s):  
Remote Sensing Data ◽  
Field Measurements ◽  
Support Vector ◽  
Lidar Data ◽  
Active Sensors ◽  
Natural Hazard Mitigation ◽  
Full Waveform ◽  
Landcover Classification ◽  
Waveform Lidar ◽  
Full Waveform Lidar

Abstract. This article presents the use of new remote sensing data acquired from airborne full-waveform lidar systems. They are active sensors which record altimeter profiles. This paper introduces a set of methodologies for processing these data. These techniques are then applied to a particular landscape, the badlands, but the methodologies are designed to be applied to any other landscape. Indeed, the knowledge of an accurate topography and a landcover classification is a prior knowledge for any hydrological and erosion model. Badlands tend to be the most significant areas of erosion in the world with the highest erosion rate values. Monitoring and predicting erosion within badland mountainous catchments is highly strategic due to the arising downstream consequences and the need for natural hazard mitigation engineering. Additionaly, beyond the altimeter information, full-waveform lidar data are processed to extract intensity and width of echoes. They are related to the target reflectance and geometry. Wa will investigate the relevancy of using lidar-derived Digital Terrain Models (DTMs) and to investigate the potentiality of the intensity and width information for 3-D landcover classification. Considering the novelty and the complexity of such data, they are presented in details as well as guidelines to process them. DTMs are then validated with field measurements. The morphological validation of DTMs is then performed via the computation of hydrological indexes and photo-interpretation. Finally, a 3-D landcover classification is performed using a Support Vector Machine classifier. The introduction of an ortho-rectified optical image in the classification process as well as full-waveform lidar data for hydrological purposes is then discussed.

scholarly journals Terrain surfaces and 3-D landcover classification from small footprint full-waveform lidar data: application to badlands

2009 ◽  
Vol 13 (8) ◽  
pp. 1531-1544 ◽  
Author(s):  
F. Bretar ◽  
A. Chauve ◽  
J.-S. Bailly ◽  
C. Mallet ◽  
A. Jacome
Keyword(s):  
Remote Sensing Data ◽  
Support Vector ◽  
Lidar Data ◽  
Digital Terrain ◽  
Natural Hazard Mitigation ◽  
Full Waveform ◽  
Mountainous Catchments ◽  
Landcover Classification ◽  
Waveform Lidar ◽  
Full Waveform Lidar

Abstract. This article presents the use of new remote sensing data acquired from airborne full-waveform lidar systems for hydrological applications. Indeed, the knowledge of an accurate topography and a landcover classification is a prior knowledge for any hydrological and erosion model. Badlands tend to be the most significant areas of erosion in the world with the highest erosion rate values. Monitoring and predicting erosion within badland mountainous catchments is highly strategic due to the arising downstream consequences and the need for natural hazard mitigation engineering. Additionally, beyond the elevation information, full-waveform lidar data are processed to extract the amplitude and the width of echoes. They are related to the target reflectance and geometry. We will investigate the relevancy of using lidar-derived Digital Terrain Models (DTMs) and the potentiality of the amplitude and the width information for 3-D landcover classification. Considering the novelty and the complexity of such data, they are presented in details as well as guidelines to process them. The morphological validation of DTMs is then performed via the computation of hydrological indexes and photo-interpretation. Finally, a 3-D landcover classification is performed using a Support Vector Machine classifier. The use of an ortho-rectified optical image in the classification process as well as full-waveform lidar data for hydrological purposes is finally discussed.

scholarly journals FWNet: Semantic Segmentation for Full-Waveform LiDAR Data Using Deep Learning

Sensors ◽  
2020 ◽  
Vol 20 (12) ◽  
pp. 3568 ◽  
Author(s):  
Takayuki Shinohara ◽  
Haoyi Xiu ◽  
Masashi Matsuoka
Keyword(s):  
Deep Learning ◽  
Semantic Segmentation ◽  
Point Clouds ◽  
Lidar Data ◽  
Global Features ◽  
Waveform Data ◽  
Full Waveform ◽  
3D Point Clouds ◽  
Waveform Lidar ◽  
Full Waveform Lidar

In the computer vision field, many 3D deep learning models that directly manage 3D point clouds (proposed after PointNet) have been published. Moreover, deep learning-based-techniques have demonstrated state-of-the-art performance for supervised learning tasks on 3D point cloud data, such as classification and segmentation tasks for open datasets in competitions. Furthermore, many researchers have attempted to apply these deep learning-based techniques to 3D point clouds observed by aerial laser scanners (ALSs). However, most of these studies were developed for 3D point clouds without radiometric information. In this paper, we investigate the possibility of using a deep learning method to solve the semantic segmentation task of airborne full-waveform light detection and ranging (lidar) data that consists of geometric information and radiometric waveform data. Thus, we propose a data-driven semantic segmentation model called the full-waveform network (FWNet), which handles the waveform of full-waveform lidar data without any conversion process, such as projection onto a 2D grid or calculating handcrafted features. Our FWNet is based on a PointNet-based architecture, which can extract the local and global features of each input waveform data, along with its corresponding geographical coordinates. Subsequently, the classifier consists of 1D convolutional operational layers, which predict the class vector corresponding to the input waveform from the extracted local and global features. Our trained FWNet achieved higher scores in its recall, precision, and F1 score for unseen test data—higher scores than those of previously proposed methods in full-waveform lidar data analysis domain. Specifically, our FWNet achieved a mean recall of 0.73, a mean precision of 0.81, and a mean F1 score of 0.76. We further performed an ablation study, that is assessing the effectiveness of our proposed method, of the above-mentioned metric. Moreover, we investigated the effectiveness of our PointNet based local and global feature extraction method using the visualization of the feature vector. In this way, we have shown that our network for local and global feature extraction allows training with semantic segmentation without requiring expert knowledge on full-waveform lidar data or translation into 2D images or voxels.

Sign in / Sign up

Export Citation Format

Share Document