Airborne Laser Scanning Point Clouds Research Articles

Building a Fully-Automatized Active Learning Framework for the Semantic Segmentation of Geospatial 3D Point Clouds

In recent years, significant progress has been made in developing supervised Machine Learning (ML) systems like Convolutional Neural Networks. However, it’s crucial to recognize that the performance of these systems heavily relies on the quality of labeled training data. To address this, we propose a shift in focus towards developing sustainable methods of acquiring such data instead of solely building new classifiers in the ever-evolving ML field. Specifically, in the geospatial domain, the process of generating training data for ML systems has been largely neglected in research. Traditionally, experts have been burdened with the laborious task of labeling, which is not only time-consuming but also inefficient. In our system for the semantic interpretation of Airborne Laser Scanning point clouds, we break with this convention and completely remove labeling obligations from domain experts who have completed special training in geosciences and instead adopt a hybrid intelligence approach. This involves active and iterative collaboration between the ML model and humans through Active Learning, which identifies the most critical samples justifying manual inspection. Only these samples (typically ≪1%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\ll 1{\\%}$$\\end{document} of Passive Learning training points) are subject to human annotation. To carry out this annotation, we choose to outsource the task to a large group of non-specialists, referred to as the crowd, which comes with the inherent challenge of guiding those inexperienced annotators (i.e., “short-term employees”) to still produce labels of sufficient quality. However, we acknowledge that attracting enough volunteers for crowdsourcing campaigns can be challenging due to the tedious nature of labeling tasks. To address this, we propose employing paid crowdsourcing and providing monetary incentives to crowdworkers. This approach ensures access to a vast pool of prospective workers through respective platforms, ensuring timely completion of jobs. Effectively, crowdworkers become human processing units in our hybrid intelligence system mirroring the functionality of electronic processing units.

AN ANALYSIS OF NEIGHBOURHOOD TYPES FOR POINTNET++ IN SEMANTIC SEGMENTATION OF AIRBORNE LASER SCANNING DATA

Abstract. The objective of the study is to conduct a comprehensive examination of how different neighbourhood types, namely spherical, cylindrical, and k-nearest neighbour (kNN), influence the feature extraction capabilities of the PointNet++ architecture in the semantic segmentation of Airborne Laser Scanning (ALS) point clouds. Two datasets are utilized for semantic segmentation analysis: the Dayton Annotated LiDAR Earth Scan (DALES) and the ISPRS 3D Semantic Labelling Benchmark datasets. In the experiments, the kNN method exhibited approximately 1% higher accuracy in weighted mean F1 and intersection over union (IoU) metrics compared to the spherical and cylindrical neighbourhood types on the DALES dataset. However, in the generalization experiment conducted on the ISPRS dataset, the spherical neighbourhood achieved the best results in these metrics, outperforming the cylindrical neighbourhood by a small margin. Notably, the kNN method was the least accurate, with a decrease in accuracy of approximately 1% in both weighted mean IoU and F1 scores. These findings suggest that the features extracted from spherical and cylindrical neighbourhood types are more generalizable compared to those from the kNN method.

Estimating tree species composition from airborne laser scanning data using point-based deep learning models

Airborne laser scanning (ALS) data is increasingly being used for accurate estimations of forest inventory attributes, such as tree height and volume. However, determining tree species information, an important attribute required by inventories for a range of forest management applications, is challenging to extract from ALS data. This is due to complex species assemblages and vertical structures in certain forest environments, and a lack of spectral information for most ALS sensors. Our objective was to estimate tree species compositions in a Canadian boreal forest environment using ALS data and point-based deep learning techniques. To accomplish this, we utilised existing polygonal forest resource information to develop a large comprehensive dataset of tree species compositions across a 630,000 ha forest management area. We then used an adapted PointAugment generative adversarial network (GAN) coupled with the dynamic graph convolutional neural network to estimate tree species proportions. This innovative deep learning approach resulted in an overall R2adj of 0.61 for all tree species proportions. A weighted F1 score of 63% was observed when classifying the leading species of a plot, 66% for leading genus, and 85% when distinguishing between coniferous and deciduous dominated plots. Furthermore, the PointAugment GAN successfully generated augmented point clouds of forest plots which can alleviate issues associated with limited training samples for use in deep learning. This approach demonstrates the capability of point-based deep learning techniques to accurately estimate tree species compositions from ALS point clouds within an expansive forested region, characterized by a complex management history and a diversity of tree species. The source code along with the pretrained models are available at https://github.com/Brent-Murray/point-dl/tree/main/Pytorch/models/PointAugment.

Enhancing LiDAR-UAS Derived Digital Terrain Models with Hierarchic Robust and Volume-Based Filtering Approaches for Precision Topographic Mapping

Airborne Laser Scanning (ALS) point cloud classification in ground and non-ground points can be accurately performed using various algorithms, which rely on a range of information, including signal analysis, intensity, amplitude, echo width, and return number, often focusing on the last return. With its high point density and the vast majority of points (approximately 99%) measured with the first return, filtering LiDAR-UAS data proves to be a more challenging task when compared to ALS point clouds. Various algorithms have been proposed in the scientific literature to differentiate ground points from non-ground points. Each of these algorithms has advantages and disadvantages, depending on the specific terrain characteristics. The aim of this research is to obtain an enhanced Digital Terrain Model (DTM) based on LiDAR-UAS data and to qualitatively and quantitatively compare three filtering approaches, i.e., hierarchical robust, volume-based, and cloth simulation, on a complex terrain study area. For this purpose, two flights over a residential area of about 7.2 ha were taken at 60 m and 100 m, with a DJI Matrice 300 RTK UAS, equipped with a Geosun GS-130X LiDAR sensor. The vertical and horizontal accuracy of the LiDAR-UAS point cloud, obtained via PPK trajectory processing, was tested using Check Points (ChPs) and manually extracted features. A combined approach for ground point classification is proposed, using the results from a hierarchic robust filter and applying an 80% slope condition for the volume-based filtering result. The proposed method has the advantage of representing with accuracy man-made structures and sudden slope changes, improving the overall accuracy of the DTMs by 40% with respect to the hierarchical robust filtering algorithm in the case of a 60 m flight height and by 28% in the case of a 100 m flight height when validated against 985 ChPs.

Modeling instream temperature from solar insolation under varying timber harvesting intensities using RPAS laser scanning

Stream temperatures are influenced by the amount of solar insolation they receive. Increasing stream temperatures associated with climate warming pose detrimental health risks to freshwater ecosystems. In British Columbia (BC), Canada, timber harvesting along forested streams is managed using riparian buffer zones of varying widths and designations. Within buffer zones, depending on distance from the stream, selective thinning may be permitted or harvest may be forbidden. In this study, we used airborne laser scanning (ALS) point cloud data acquired via a remotely piloted aircraft system (RPAS) to derive forest canopy characteristics that were then used to estimate daily incoming summer and fall solar insolation for five stream reaches in coastal conifer-dominated temperate forests in Vancouver Island, BC, Canada. We then examined empirical relationships between estimated insolation and actual instream temperature measurements. Based on these empirical relationships, the potential effects of timber harvest on instream temperatures were simulated by comparing scenarios of different riparian forest harvest intensities. Our results indicated that modeled solar insolation explained 43–90 % of the variation in observed stream reach temperatures, and furthermore, when a single cold-water stream reach was excluded explained an overall 81 % of variation. Simulated harvesting scenarios generally projected increases in maximum stream reach temperatures 1–2 °C in summer and early fall months. However, in a full clearcut scenario (i.e. where all trees were removed), maximum stream reach temperatures increased as much as 5.8 °C. Our results emphasize the importance of retaining riparian vegetation for the maintenance of habitable temperatures for freshwater-reliant fish with thermal restrictions. In addition, we demonstrate the feasibility of RPAS-based monitoring of stream reach shading and canopy cover, enabling detailed assessment of environmental stressors faced by fish populations under climate warming.

Carbon Losses from Topsoil in Abandoned Peat Extraction Sites Due to Ground Subsidence and Erosion

Peat erosion has a significant impact on soil fertility, agricultural productivity, and climate change dynamics. Through this process, the topsoil rich in organic matter and carbon (C) is removed and can travel long distances, causing a net C loss. Additionally, peat undergoes oxidation, resulting in further C loss. In our study, we evaluated C losses from 11 peat extraction fields in two study sites, abandoned for more than 15 years and overgrown by vegetation of different densities. We used high-resolution airborne laser scanning point clouds and multispectral aerial images acquired periodically within a 9-year period, as well as chemical analyses of the topsoil layer. In our study, we found a strong correlation between peat subsidence, C loss, and the vegetation density (NDVI value). NDVI also determines most of the uncertainty in elevation data. We found also that both erosion and peat subsidence are significant sources of C losses from peat extraction sites. At a site monitored for over 9 years, our estimated ground elevation changes ranged from 0.1 cm y−1 to 0.58 cm y−1; however, at a different site monitored over a 4-year period, the values ranged from 2.14 cm y−1 to 5.72 cm y−1. Accordingly, the mean annual C losses varied from 0.06 to 0.22 kg C m−2 y−1 and from 1.21 to 3.57 kg C m−2 y−1.

Classification of tree forms in aerial LiDAR point clouds using CNN for 3D tree modelling

ABSTRACT Three-dimensional models of trees that correspond to the real-world forms of the trees on the ground are used in urban planning, solar power estimation, and other disciplines. Previous studies have focused on generating 3D tree models from high-density point cloud data such as Terrestrial Laser Scanning (TLS) data, which is expensive and limited to small spatial extents. However, there has been limited exploration of inexpensive solutions to model trees over large spatial extents. The goal of this study is to use widely available discrete return Airborne Laser Scanning (ALS) data along with field-captured tree photographs and Google Street View (GSV) images to develop 3D equivalents of trees over larger spatial extents. To this end, we designed a process to assign representative 3D models for individual trees in discrete return ALS point clouds. This study demonstrates the use of a Convolutional Neural Network (CNN) model and 3D models generated with Structure from Motion (SfM) for the realistic modelling of deciduous non-overlapping trees from discrete return ALS data. We classified and labelled the crown shapes of deciduous trees in a study area into four classes based on GSV images of trees. We delineated and segmented non-overlapping deciduous trees from ALS data and reduced them to 2D images using voxel point counts. Next, we trained a CNN architecture to match the 2D images to the corresponding classes observed from GSV images. For each class, we created a representative 3D tree model using field-captured circumnavigational photos of trees and SfM. To demonstrate 3D visualization using the 3D tree models, we created a 3D visualization of the trees surrounding a parking lot. The trained CNN model from this study can be used to classify non-overlapping deciduous trees from discrete return ALS data and subsequently visualize near-realistic 3D tree models of trees.

Assessing Forest Traversability for Autonomous Mobile Systems Using Close-Range Airborne Laser Scanning

Advances in sensor technology and computing performance has brought us into an era of digital forestry where a forest environment can be digitally replicated. At the same time, an increasing interest in the use of unmanned vehicles and other autonomous mobile systems (AMSs) in forest mapping and operations has emerged. However, a forest is an unstructured and rather complex environment for AMSs to operate in, and usually some kind of a priori information of traversability is required. The aim of this study was to assess forest traversability for AMSs using high-density airborne laser scanning (ALS) point clouds. It was assumed that such point clouds acquired from a helicopter flying at a low altitude can be used to characterise vegetation obstacles affecting forest traversability. A voxel-based vegetation occupancy analysis was carried out with the aim to detect open space to define traversable three-dimensional space. The experimental setup included seven sample plots (32×32 m) representing diverse boreal forest structures. Terrestrial laser scanning (TLS) was used for obtaining reference for vegetation occupancy. Comparison between ALS and TLS revealed an overall accuracy of 0.85–0.94 with a recall of 0.78–0.91 and a precision of 0.62–0.74 for ALS-based voxel classification for vegetation occupancy depending on forest structure. This implies that up to 91% of voxels assigned a classification »occupied« based on the TLS could be correctly classified using the ALS, while up to 74% of voxels assigned a classification »occupied« using the ALS were occupied based on the TLS. Density of low vegetation accounted for 83% of the variation in accuracy and precision. The feasibility of vegetation occupancy information to be used by an AMS for navigation was also demonstrated. It was assumed that the ALS data convey as sufficient information of AMS path planning as does the TLS data. The experiments showed that out of 1393 randomly generated paths based on empty space detected by the ALS, 72% were considered feasible when validated with the TLS data. The success rate in path planning varied from 0.54 to 0.92 between the sample plots and was seemingly affected by vegetation density that accounted for 53% of variation in success rate. Altogether, the demonstrated possibility to predefine forest traversability using remote sensing will support the use of AMSs in forestry.

How to adequately determine the top height of forest stands based on airborne laser scanning point clouds?

Top height (TH) is one of the most important features of forest stands that is used in forest resource management. Thus far, the TH has been mainly determined in forestry on the basis of field measurements. There are certain accepted theoretical guidelines on how the TH should be determined. However, due to high cost and time consumption, it is difficult to meet these assumptions with traditional field measurement methods. Currently, remote sensing is widely used in forest inventory, also for the determination of TH. However, it is still usually conducted by linking the remote sensing data—most often airborne laser scanning (ALS) point clouds—with ground-based measurements using the area-based approach. Different plot sizes and point cloud metrics are used to determine TH, making it difficult to compare the results. There is therefore a need for a universal methodology for determining TH directly from ALS data – one that is consistent with the methodology used for field measurement data. This paper presents the results of an experiment that was aimed at assessing the accuracy of TH determination on the basis of ALS data. A comparison of the six considered ALS-derived proxies of TH showed that the individual tree detection approach was the most accurate. The bias of this method, when calculated for 48 artificial forest stands, amounted to 0.15 m. The approach based on the lower percentiles of the distribution of point heights significantly underestimated the TH from −0.67 m to up to −1.9 m. On the basis of the obtained results, we claim that TH can be determined, based on ALS data without ground measurements, with an accuracy that is acceptable for forest practitioners and researchers.

IPCONV: Convolution with Multiple Different Kernels for Point Cloud Semantic Segmentation

The segmentation of airborne laser scanning (ALS) point clouds remains a challenge in remote sensing and photogrammetry. Deep learning methods, such as KPCONV, have proven effective on various datasets. However, the rigid convolutional kernel strategy of KPCONV limits its potential use for 3D object segmentation due to its uniform approach. To address this issue, we propose an Integrated Point Convolution (IPCONV) based on KPCONV, which utilizes two different convolution kernel point generation strategies, one cylindrical and one a spherical cone, for more efficient learning of point cloud data features. We propose a customizable Multi-Shape Neighborhood System (MSNS) to balance the relationship between these convolution kernel point generations. Experiments on the ISPRS benchmark dataset, LASDU dataset, and DFC2019 dataset demonstrate the validity of our method.

Performance, effectiveness and computational efficiency of powerline extraction methods for quantifying ecosystem structure from light detection and ranging

ABSTRACT National and regional data products of the ecosystem structure derived from airborne laser scanning (ALS) surveys with Light Detection And Ranging (LiDAR) technology are essential for ecology, biodiversity, and ecosystem monitoring. However, noises like powerlines often remain, hindering the accurate measurement of 3D ecosystem structures from LiDAR. Currently, there is a lack of studies assessing powerline noise removal in the context of generating data products of ecosystem structures from ALS point clouds. Here, we assessed the (1) performance and accuracy, (2) effectiveness, and (3) time efficiency and execution time of three powerline extraction methods (i.e. two point-based methods based on deep learning and eigenvalue decomposition, respectively, and one hybrid method) for removing powerline noise when quantifying 3D ecosystem structures in landscapes with varying canopy heights and vegetation openness. Twenty-five LiDAR metrics representing three key dimensions of the ecosystem structure (i.e. vegetation height, cover, and vertical variability) across 10 study areas in the Netherlands were used for our assessment. The deep learning method had the best performance and showed the highest accuracy of powerline removal across various landscape types (average F 1 score = 96%), closely followed by the hybrid method (average F 1 score = 95%). In contrast, the accuracy of the eigenvalue decomposition method was lower (average F 1 score = 82%) and depended on landscape context and vegetation composition (e.g. the F 1 score decreased from 96% to 63% when the average canopy height increased across landscapes). Powerline noise removal had the highest effectiveness (i.e. generating LiDAR metrics closest to those derived from manually labeled ground truth data) for LiDAR metrics capturing height and cover of low- and high-vegetation layers. Time efficiency (processed points per second) was highest for the eigenvalue decomposition method, yet the hybrid method reduced the execution time by > 50% compared to the deep learning method (ranging from 20% to 89% in study areas with different landscape composition). Based on our findings, we recommend the hybrid method for upscaling powerline removal on multi-terabyte ALS datasets to a regional or national extent because of its high accuracy and computational efficiency. Remaining misclassifications in LiDAR metrics could be further minimized by improving the training dataset for deep learning models (e.g. including various shapes of transmission towers from different datasets). Our findings provide novel insights into the performance of different powerline extraction methods, how their effectiveness varies for improving vegetation metrics and mapping the 3D ecosystem structure from LiDAR, and their computational efficiency for upscaling powerline removal in multi-terabyte ALS datasets to a national extent.

Which metrics derived from airborne laser scanning are essential to measure the vertical profile of ecosystems?

AbstractIn a recent perspective (Diversity and Distributions, 29, 39–50), ‘10 variables’ were proposed to measure vegetation structure from airborne laser scanning (ALS) for assessing species distributions and habitat suitability. We worry about this list because the variables predominantly represent variation in vegetation height, the vertical variability of vegetation biomass is insufficiently captured, and variables of vegetation cover are ill‐defined or not ecosystem agnostic. We urge for a better defined, more comprehensive and more balanced list, and for assessing which information from ALS point clouds is truly essential to measure the major dimensions of 3D vegetation structure within and across ecosystems and animal habitats. We think that the currently proposed ‘list of 10 ALS metrics’ is premature and that researchers and stakeholders should be cautious in adopting this list.

Improving Estimation of Tree Parameters by Fusing ALS and TLS Point Cloud Data Based on Canopy Gap Shape Feature Points

Airborne laser scanning (ALS) and terrestrial laser scanning (TLS) are two ways to obtain forest three-dimensional (3D) spatial information. Due to canopy occlusion and the features of different scanning methods, some of the forest point clouds acquired by a single scanning platform may be missing, resulting in an inaccurate estimation of forest structure parameters. Hence, the registration of ALS and TLS point clouds is an alternative for improving the estimation accuracy of forest structure parameters. Currently, forest point cloud registration is mainly conducted based on individual tree attributes (e.g., location, diameter at breast height, and tree height), but the registration is affected by individual tree segmentation and is inefficient. In this study, we proposed a method to automatically fuse ALS and TLS point clouds by using feature points of canopy gap shapes. First, the ALS and TLS canopy gap boundary vectors were extracted by the canopy point cloud density model, and the turning or feature points were obtained from the canopy gap vectors using the weighted effective area (WEA) algorithm. The feature points were then aligned, the transformation parameters were solved using the coherent point drift (CPD) algorithm, and the TLS point clouds were further aligned using the recovery transformation matrix and refined by utilizing the iterative closest point (ICP) algorithm. Finally, individual tree segmentations were performed to estimate tree parameters using the TLS and fusion point clouds, respectively. The results show that the proposed method achieved more accurate registration of ALS and TLS point clouds in four plots, with the average distance residuals of coarse and fine registration of 194.83 cm and 2.14 cm being much smaller compared with those from the widely used crown feature point-based method. Using the fused point cloud data led to more accurate estimates of tree height than using the TLS point cloud data alone. Thus, the proposed method has the potential to improve the registration of ALS and TLS point cloud data and the accuracy of tree height estimation.

Mapping and Estimating Aboveground Biomass in an Alpine Treeline Ecotone under Model-Based Inference

Due to climate change, treelines are moving to higher elevations and latitudes. The estimation of biomass of trees and shrubs advancing into alpine areas is necessary for carbon reporting. Remotely sensed (RS) data have previously been utilised extensively for the estimation of forest variables such as tree height, volume, basal area, and aboveground biomass (AGB) in various forest types. Model-based inference is found to be efficient for the estimation of forest attributes using auxiliary RS data, and this study focused on testing model-based estimations of AGB in the treeline ecotone using an area-based approach. Shrubs (Salix spp., Betula nana) and trees (Betula pubescens ssp. czerepanovii, Sorbus aucuparia, Populus tremula, Pinus sylvestris, Picea abies) with heights up to about five meters constituted the AGB components. The study was carried out in a treeline ecotone in Hol, southern Norway, using field plots and point cloud data obtained from airborne laser scanning (ALS) and digital aerial photogrammetry (DAP). The field data were acquired for two different strata: tall and short vegetation. Two separate models for predicting the AGB were constructed for each stratum based on metrics calculated from ALS and DAP point clouds, respectively. From the stratified predictions, mean AGB was estimated for the entire study area. Despite the prediction models showing a weak fit, as indicated by their R2-values, the 95% CIs were relatively narrow, indicating adequate precision of the AGB estimates. No significant difference was found between the mean AGB estimates for the ALS and DAP models for either of the strata. Our results imply that RS data from ALS and DAP can be used for the estimation of AGB in treeline ecotones.

Deep learning for filtering the ground from ALS point clouds: A dataset, evaluations and issues

The capability of partially penetrating vegetation canopy and efficiently collecting high-precision point clouds over large areas makes airborne laser scanning (ALS) a valuable tool for various geospatial applications. However, automated ground filtering (GF), one fundamental and challenging step for most ALS applications, has remained a widely researched yet unsolved problem for decades. The recent breakthroughs in supervised deep learning (DL) techniques, which rely on sufficient and high-quality labeled datasets, provide a new solution to better solve this problem. Unfortunately, public 3D geospatial datasets are scarce, especially for those tailored for the landform-scale GF task. Moreover, whether advanced deep neural networks (DNNs) can be well-scaled to the problem of GF remains an open question. To comprehensively advance the development of effective DL-based GF pipelines, we first publish an ultra-large-scale GF dataset built upon open-access ALS point clouds of four different countries worldwide, which covers over 47 km2 and nine different terrain scenes. Then, multiple attractive advantages of DL techniques in GF are evaluated through extensive experimental comparisons with traditional GF methods on the presented dataset. Furthermore, we reveal several issues faced by generalizing existing advanced 3D DNNs into GF tasks with a series of in-depth experimental analyses. Finally, some promising directions for future research are suggested in response to the identified challenges. Our dataset, named OpenGF, is available at https://github.com/Nathan-UW/OpenGF.

DGCN-ED: dynamic graph convolutional networks with encoder–decoder structure and its application for airborne LiDAR point classification

ABSTRACT Achieving high-precision automatic classification in real-world applications for airborne laser scanning point clouds is a challenging task duo to their unstructured nature, uneven density distribution, high redundancy, incompleteness and scene complexity. Graph Convolutional Neural Networks can process scattered point clouds directly without regularization, which avoids the loss of depth information，and has recently become a topic of increased interest. Therefore, this study proposed an extension of the Graph-Unet network architecture named DGCN-ED for airborne LiDAR point classification, which uses a Graph Convolutional Neural Network as a representation to describe complex object relationships and an encoder-decoder architecture to capture the multi-scaled point features and describe objects in the high-level feature space. The two-layer dynamic update Graph Convolutional Neural Network is designed to expand the effective range of nodes and enhance the representation ability of learned pointwise features. The effectiveness of the proposed method is evaluated by an experiment on the ISPRS Vaihingen 3D semantic labelling benchmark dataset. Moreover, experiments on the IEEE 2019 Data Fusion Contest Dataset were conducted to demonstrate the generalization abilities of the proposed method. The results show that our method achieved on average 2.8% higher overall accuracy than existing methods, with an overall accuracy of 98.0% and an average F1 score of 0.797.

Detecting the presence of standing dead trees using airborne laser scanning and optical data

ABSTRACT Deadwood is an important indicator of biodiversity in forest ecosystems. Identifying areas with large density of standing dead trees through field inventory is challenging, and remotely sensed data can provide a more systematic approach. In this study, we used metrics derived from airborne laser scanning (ALS) data (7.1 points m−2) and vegetation indices from optical images (HySpex sensor VNIR-1800: 0.3 m, SWIR-384: 0.7 m) to predict the presence of standing dead trees over a 15.9 km2 managed forest in Southern Norway. The dead basal area (DBA) of 40 sample plots was computed and used to classify the plots into presence/absence of standing dead trees. An area-based approach (ABA) using logistic regression was initially tested, but due to limited ground reference information, no statistically significant models could be formulated. A tree-based approach (TBA) was used to overcome this limitation. It identified trees on the ALS point cloud with a local maxima function and used a vegetation index to determine if the trees were dead. Between 18% and 42% of the predicted area with standing dead trees intersected a field recorded validation dataset. The TBA provided a good alternative to area-based regression models in the context of few standing dead trees.

Instance recognition of street trees from urban point clouds using a three-stage neural network

As one of the most important components of urban space, the geometric and semantic properties of road trees are crucial for the assessment and upgrade of urban environments. Accurate positioning of individual trees is a critical procedure for this task; however, traditional manual methods require considerable personnel and material resources. By contrast, individual tree segmentation based on a vehicle-mounted mobile laser scanner (MLS) operating along urban streets can provide a feasible and efficient solution. Recent studies (especially those based on deep learning) on point cloud semantic segmentation and forest tree detection have achieved significant progress; however, they rarely perform instance recognition of urban trees. In addition, extracting individual urban trees from point clouds to balance accuracy and efficiency remains challenging. To this end, we propose an effective pipeline to segment individual trees from urban MLS data. The proposed approach encompasses two main steps: tree-point extraction and individual-tree segmentation. First, a patch-guided semi-supervised semantic segmentation network was designed to separate non-tree points from entire urban scene. The extracted tree points are further divided into instance-level trees using a three-stage segmentation network for the individual tree segmentation. Specifically, a tree-centroid prediction module is proposed to ensure the accurate localization of street trees, even with irregular positions, in complex urban environments. Subsequently, a special confidence-aware attention mechanism is developed to resolve challenging ambiguities from complementary perspectives. Finally, a dynamic shifting-based clustering algorithm is introduced to speed up the segmentation of each tree guided by the predicted centroid and directly transfer the IDs back to the point clouds. The proposed method is evaluated and validated on MLS point clouds collected in two urban areas, robustly producing accurate instance-level roadside trees. The instance-level segmentation results in average precision and recall of (96.62%, 96.17%) and (95.62%, 94.58%), respectively, on these two MLS point cloud datasets. Point-level segmentation also exhibits mean precision, recall, and weighted coverage exceeding 90%. We also verify that the proposed method could achieve a promising result on an airborne laser scanning point cloud dataset without retraining, thereby demonstrating a good generalization ability. To further verify the quality of the segmented tree instances, a reconstruction method is adopted to employ these independent street trees in trunk point cloud modeling. Our contribution may assist in developing intelligent urban green space planning and management.

An Optimized Workflow for Digital Surface Model Series Generation Based on Historical Aerial Images: Testing and Quality Assessment in the Beach-Dune System of Sa Ràpita-Es Trenc (Mallorca, Spain)

We propose an optimized Structure-from-Motion (SfM) Multi-View Stereopsis (MVS) workflow, based on minimizing different errors and inaccuracies of historical aerial photograph series (1945, 1979, 1984, and 2008 surveys), prior to generation of elevation-calibrated historical Digital Surface Models (hDSM) at 1 m resolution. We applied LiDAR techniques on Airborne Laser Scanning (ALS) point clouds (Spanish PNOA LiDAR flights of 2014 and 2019) for comparison and validation purposes. Implementation of these products in multi-temporal analysis requires quality control due to the diversity of sources and technologies involved. To accomplish this, (i) we used the Mean Absolute Error (MAE) between GNSS-Validation Points and the elevations observed by DSM-ALS to evaluate the elevation accuracy of DSM-ALS generated with the LAScatalog processing engine; (ii) optimization of the SfM sparse clouds in the georeferencing step was evaluated by calculating the Root Mean Square Error (RMSE) between the Check Points extracted from DSM-ALS and the predicted elevations per sparse cloud; (iii) the MVS clouds were evaluated by calculating the MAE between ALS-Validation Points and the predicted elevations per MVS cloud; iv) the accuracy of the resulting historical SfM-MVS DSMs were assessed using the MAE between ALS-Validation Points and the observed elevations per historical DSM; and (v) we implemented a calibration method based on a linear correction to reduce the elevation discrepancies between historical DSMs and the DSM-ALS 2019 reference elevations. This optimized workflow can generate high-resolution (1 m pixel size) hDSMs with reasonable accuracy: MAE in z ranges from 0.41 m (2008 DSM) to 5.21 m (1945 DSM). Overall, hDSMs generated using historical images have great potential for geo-environmental processes monitoring in different ecosystems and, in some cases (i.e., sufficient image overlapping and quality), being an acceptable replacement for LiDAR data when it is not available.

Vectorial and topologically valid segmentation of forestry road networks from ALS data

Accurate information on road location is critical for forest management and conservation strategies. Road location data supports the analysis of road accessibility and usability and is a critical information layer for forest harvest, financial planning, wildfire suppression, and protection activities. The global expanse of forests, their remoteness, and difficulty to access have necessitated the development of automatic or semi-automatic remote sensing methodologies to map roads using passive optical imagery or Airborne Laser Scanning (ALS). Conventional automatic road mapping methods are raster-based and map roads as patches of disconnected pixels. This paper addresses the limitations of raster-based automatic forest road extraction and presents a method for producing a topologically accurate vectorial road network. Our method, presented as a fully documented and open-source software tool, uses metrics derived from an ALS point cloud to produce a raster of road conductivity. From this conductivity raster, the method “drives” the roads iteratively by detecting and following road intersections. We demonstrate the method’s efficacy using a road network in Quebec, Canada, where 96% of the roads in a binary raster, and 84% using our probability map, are vectorized properly from an ALS point cloud with 4% false positives. Our proposed method may significantly reduce the training requirements of machine learning techniques used to classify roads by being very robust to false positive and false negative classifications.