Airborne LiDAR Point Clouds Research Articles

Research on Deep Learning Individual Tree Segmentation Method Coupling RetinaNet and Point Cloud Clustering

Increasing human activities have caused serious disturbance to global forest resources, so how to accurately identify individual trees has become an important task of forest resources investigation. In order to get the accurate number of individual trees, this paper took coniferous forest and mixed coniferous and broad-leaved forest as experimental samples, as well as Digital Orthophoto Map and airborne LiDAR Point Cloud as research data. We propose a deep Learning individual tree segmentation method based on RetinaNet model and PCS algorithm by doing comparative analysis (classical Watershed Algorithm and Layer Stacking Algorithm) at the plots (with high, medium, and low densities). The experimental results show that the method proposed in this paper can solve the problem of individual tree segmentation in high density forest and improve its degree of automation. Compared with Watershed algorithm and Layer stacking algorithm, F-Measure is improved by 6%-29% and 7%-20%, respectively. In other words, the results of individual tree segmentation presented in this paper can not only improve the precision of individual tree segmentation, but also maintain a high detection rate, which can meet the accuracy and high efficiency of individual tree extraction, so as to meet the needs for modern forestry investigation.

Slope extraction algorithm of ski tracks based on airborne LIDAR point cloud

Slope extraction algorithm of ski tracks based on airborne LIDAR point cloud

Optimal Model Fitting for Building Reconstruction From Point Clouds

Geometric-semantic coherent building models are demanding in many geoscience applications. Conventional building modeling methods often rely on successive roof plane segmentation and fitting. The subsequent reconstruction procedure is difficult to assure topologic consistency and geometric accuracy. This article starts with a library of predefined building models or primitives, including pyramid, gable, hip, etc. We propose an optimal model fitting approach that holistically determines all of its parameters from segmented point cloud data. The approach is formulated as an optimization problem that minimizes the point-to-mesh distance between the point cloud and the meshed primitive model. Necessary constraints in the form of inequality equations are introduced to assure correct and reliable solution. For complex roofs consisting of several predefined primitive models, a hierarchical procedure is presented to reconstruct the major roof model and its superstructures sequentially. The CityGML LoD2 model is created from the parameterized primitives. The quality and performance of this approach are evaluated with airborne lidar and photogrammetric point clouds. Based on the experiments with 910 buildings, the primitive fitting accuracy is 7.8 cm and the corner uncertainty is 0.36 m or 0.78 times the ground point spacing; the building boundary consistency is 89.6%. The study demonstrates a piecewise continuous polyhedral building model can be determined through a holistic parameter optimization process. The resultant building models intrinsically best fit to the input point cloud with topologic integrity. The approach not only qualitatively generates semantic building models but also exhibits the potential for building reconstruction over large areas.

Airborne LiDAR and Photogrammetric Point Cloud Fusion for Extraction of Urban Tree Metrics According to Street Network Segmentation

This paper provides a practical procedure for fusing LiDAR and photogrammetric point clouds for the extraction of tree metrics. Aerial photogrammetric point clouds are first generated using the structure-from-motion and dense-matching methods. Registration of the LiDAR and photogrammetric point clouds is then performed using an onboard global positioning system and inertial measurement unit. However, due to systematic deviations, it is impossible to directly merge the two types of point cloud. Therefore, an urban street network obtained from the OpenStreetMap digital mapping system is utilized for point cloud segmentation. After segmentation, each chunk is finely registered and merged based on the iterative closest point algorithm, allowing the two types of data to be accurately co-registered and a fused point cloud obtained. Finally, we conducted experiments to extract stand and individual tree metrics from fused point clouds created for two study plots. The height distributions of the fused point clouds were highly consistent with LiDAR data, with the 5%, 10%, 25%, 50%, 75%, 90%, and 95% height percentiles showing acceptable similarities. The height distribution of individual trees was also consistent with that of field measurements. Furthermore, the fused point clouds contain a high point density and RGB color information, which allow shape delineation and estimation of tree health status. This comprehensive analysis demonstrates that this procedure provides a practical way to inventory tree stands and individuals in urban areas.

Small-Sample Airborne LiDAR Point Cloud Classification Based on Transfer Learning and Fully Convolutional Network

Small-Sample Airborne LiDAR Point Cloud Classification Based on Transfer Learning and Fully Convolutional Network

Scan Line Void Filling of Airborne LiDAR Point Clouds for Hydroflattening DEM

Generation of LiDAR-derived digital elevation model (DEM), particularly for hydrologic and shore environments, poses a continuous challenge. The presence of laser dropouts found on the water bodies causes data voids/holes in the airborne LiDAR data point clouds. Unnatural huge triangular artifacts may appear in these regions when a DEM is generated, resulting in not only unpleasant visual effect but also inaccurate terrain analyses. The United States Geological Survey has stressed the need of having a hydro-flattened DEM in the LiDAR Base Specification. Different forms of water bodies should be represented by a flat surface. Existing approaches mainly rely on the use of ancillary data or manual intervention during the hydroflattening process. In this study, an automatic data processing workflow is proposed to: 1) classify land and water data points collected by a topographic airborne LiDAR system based on the scan line intensity-elevation ratio; 2) perform scan line void filling of data points in close-to-nadir region and at both swath edges; 3) generate a virtual water surface based on the classified water data points; 4) perform hydroflattening on the DEM. The proposed workflow was examined using five datasets collected by topographic airborne LiDAR on the inland ponds and lakes, inland rivers, nontidal boundary water bodies, tidal water bodies, and islands, as addressed in the LiDAR Base Specification. The results showed that the proposed workflow can successfully generate hydroflattened DEMs and overcome the drawback of existing approaches.

A Comparison of Deep Learning Methods for Airborne Lidar Point Clouds Classification

The success achieved by deep learning techniques in image labeling has triggered a growing interest in applying deep learning for three-dimensional point cloud classification. To provide better insights into different deep learning architectures and their applications to ALS point cloud classification, this article presents a comprehensive comparison among three state-of-the-art deep learning networks: PointNet++, SparseCNN, and KPConv, on two different ALS datasets. The performances of these three deep learning networks are compared w.r.t. classification accuracy, computation time, generalization ability as well as the sensitivity to the choices of hyper-parameters. Overall, we observed that PointNet++, SparseCNN, and KPConv all outperform Random Forest on the classification results. Moreover, SparseCNN leads to a slightly better classification result compared to PointNet++ and KPConv, while requiring less computation time and memory. At the same time, it shows a better ability to generalize and is less impacted by the different choices of hyper-parameters.

Interpolation-based filtering with segmentation for airborne LiDAR point clouds

Interpolation-based filtering with segmentation for airborne LiDAR point clouds

Initial Road Extraction of Airborne LiDAR Point Cloud

Initial Road Extraction of Airborne LiDAR Point Cloud

A VOXEL-BASED METHOD TO ESTIMATE NEAR-SURFACE AND ELEVATED FUEL FROM DENSE LIDAR POINT CLOUD FOR HAZARD REDUCTION BURNING

Abstract. Drastic changes in the climate has revised the face of disaster management: it is contributing to abnormal intensity, frequency and duration of extreme weather and climate events. The year 2020 started with more than 100 fires burning across Australia. Hazard reduction burning has become a resolute and primary land management technique that contribute to the reduction of bushfire severity. One of the key variables to consider for this application is fuel load, as the accumulation of vegetation in a forest profile affects the intensity of the burn. Conventionally, fuel loads are measured by manually cutting the vegetation and physically measuring the quantity after dry heating. This process is expensive, and time consuming. There is an opportunity for these techniques to be digitised and automated to give results in a timely manner and work as a decision support tool for practitioners. This paper proposes a voxel-based approach that can be used for estimating fuel load and percentage cover of the vegetation, at the elevated and near-surface fuel/vegetation layer as a method to augment manual estimation. We use an airborne LiDAR pointcloud dataset of Vermont Place Park, Newcastle, Australia to test the method. The preliminary inspection of the results confirms the technique that can approximate conventional manual method. Next steps include performance testing including more dataset to derive quantitative measures on the approach.

Automatic extraction and reconstruction of powerlines from airborne LiDAR point clouds

Automatic extraction and reconstruction of powerlines from airborne LiDAR point clouds

Comparison of Filters for Archaeology-Specific Ground Extraction from Airborne LiDAR Point Clouds

Identifying bare-earth or ground returns within point cloud data is a crucially important process for archaeologists who use airborne LiDAR data, yet there has thus far been very little comparative assessment of the available archaeology-specific methods and their usefulness for archaeological applications. This article aims to provide an archaeology-specific comparison of filters for ground extraction from airborne LiDAR point clouds. The qualitative and quantitative comparison of the data from four archaeological sites from Austria, Slovenia, and Spain should also be relevant to other disciplines that use visualized airborne LiDAR data. We have compared nine filters implemented in free or low-cost off-the-shelf software, six of which are evaluated in this way for the first time. The results of the qualitative and quantitative comparison are not directly analogous, and no filter is outstanding compared to the others. However, the results are directly transferable to real-world problem-solving: Which filter works best for a given combination of data density, landscape type, and type of archaeological features? In general, progressive TIN (software: lasground_new) and a hybrid (software: Global Mapper) commercial filter are consistently among the best, followed by an open source slope-based one (software: Whitebox GAT). The ability of the free multiscale curvature classification filter (software: MCC-LIDAR) to remove vegetation is also commendable. Notably, our findings show that filters based on an older generation of algorithms consistently outperform newer filtering techniques. This is a reminder of the indirect path from publishing an algorithm to filter implementation in software.

Relationship between tree height and landslide characteristics obtained by GIS assessment

AbstractLandslides in forested landscapes have far‐reaching implications, beyond that of just destroying the forest itself, sometimes initiating large‐scale sediment disasters. Although vegetation increases slope stability through its root network, it is hard to evaluate its contribution to slope stability over a wide area. In this study, the relationship between tree height and landslide characteristics in the Ikawa catchment, central Japan, was investigated to develop a method for evaluating the effects of forest cover on slope stability over a regional extent. Catchment‐wide tree height was obtained using airborne LiDAR point cloud data and used in conjunction with the root depth profile, measured for trees of various height by digging trenches. Root tensile strength per unit area of soil was calculated from individual root diameters and empirical power law equations on the relationship between root diameter and root tensile force in order to better understand the effect that tree height has on slope stability. Landslide density in the Ikawa catchment shows that landslides occur more frequently in forests with shorter trees, with occurrence decreasing as tree height increases. This is likely due to the stabilizing features of larger trees having a greater network of roots, which is supported by the general increase in total root area and the deeper penetration of root biomass into the soil as the height of trees surveyed increases. Landslide density was not solely affected by tree height, but also by slope gradient and plane curvature. Decreasing landslide occurrence and landslide area as tree height increases suggests that slope stability increases with tree height, while the random distribution of results when comparing landslide depth to tree height suggests that while tree height has an impact on relative slope stability, the landslide failure depth is independent of tree height, and thus controlled by other factors. © 2020 John Wiley & Sons, Ltd.

Utilizing Airborne LiDAR and UAV Photogrammetry Techniques in Local Geoid Model Determination and Validation

This investigation evaluates the performance of digital terrain models (DTMs) generated in different vertical datums by aerial LiDAR and unmanned aerial vehicle (UAV) photogrammetry techniques, for the determination and validation of local geoid models. Many engineering projects require the point heights referring to a physical surface, i.e., geoid, rather than an ellipsoid. When a high-accuracy local geoid model is available in the study area, the physical heights are practically obtained with the transformation of global navigation satellite system (GNSS) ellipsoidal heights of the points. Besides the commonly used geodetic methods, this study introduces a novel approach for the determination and validation of the local geoid surface models using photogrammetry. The numeric tests were carried out in the Bergama region, in the west of Turkey. Using direct georeferenced airborne LiDAR and indirect georeferenced UAV photogrammetry-derived point clouds, DTMs were generated in ellipsoidal and geoidal vertical datums, respectively. After this, the local geoid models were calculated as differences between the generated DTMs. Generated local geoid models in the grid and pointwise formats were tested and compared with the regional gravimetric geoid model (TG03) and a high-resolution global geoid model (EIGEN6C4), respectively. In conclusion, the applied approach provided sufficient performance for modeling and validating the geoid heights with centimeter-level accuracy.

TERRAIN-ADAPTIVE GROUND FILTERING OF AIRBORNE LIDAR DATA BASED ON SALIENCY-AWARE THIN PLATE SPLINE

Abstract. Ground filtering separates the ground and non-ground points from point clouds, which is the essential process for DEM generation, semantic segmentation, model reconstruction and so forth. Considering the topologically complex terrain environments, the segmentation results are prone to be disturbed dealing with steep slopes, buildings, bridges, cliffs, etc. from Airborne LiDAR point clouds. In this paper, a saliency-aware Thin-Plate-Spline (SATPS) interpolation method is proposed including two steps: saliency division and adaptive regularized TPS interpolation with relative variance coefficient. Firstly, the point clouds are indexed in 2D grids and segments are clustered step probing toward 8-adjacent scanning directions. Then, the saliency of each grid is calculated according to the elevation variance of adjacent segments towards each scanning direction. Subsequently, grids of high ground saliency are considered as candidates for seed point selection and then clustered by region growing. The TPS surface is interpolated for each cluster loosely fitting to the seed points involving an adaptive relative variance coefficient which is according to ground saliency and elevation deviation. And finally, the ground points are extracted around the TPS surface. Experimental results indicate that the proposed SATPS algorithm achieves better Type 1 accuracy and total accuracy than the state-of-the-art algorithms in scenes with complex terrain structures, which is practical to generate DEM products.

QUALITY IMPROVEMENT FOR AIRBORNE LIDAR DATA FILTERING BASED ON DEEP LEARNING METHOD

Abstract. In this paper, we discuss how to improve the quality of classification results when deep learning is applied for the filtering of airborne LiDAR point cloud. We introduce the baseline method which utilizes convolutional neural network (CNN) based on voxelization, and then we propose three methods to improve the quality of classification result. The first method is data pre-processing that aims to exclude data in advance that is clearly not on the ground surface in order to efficiently extract the ground surface data. Data pre-processing can greatly reduce the number of target points and the subsequent processing can be performed efficiently. It also has the effect of preventing noise-like points floating in the air from being misclassified as the ground surface, as compared to the case without pre-processing. The second method is changing the network structure. In recent years, various networks have been proposed for classifying point clouds. In our study, the baseline is using very simple networks. In order to improve the classification result of the baseline method, the layer depth and the range size of convolution are changed, and we investigated about the improvements of the results. The current discussion can be used as a guidance when considering new networks. The third method is the integration of classification results from multiple networks. We integrated individual results from multiple networks with varying layer depths and convolution sizes, starting with the baseline, and investigated whether the results improved. We observed that even if the individual results were similar, the classification results can be improved by integrating the results.

MIN-CUT BASED SEMANTIC BUILDING LABELING FOR AIRBORNE LIDAR DATA

Abstract. Classification and segmentation of buildings from airborne lidar point clouds commonly involve point features calculated within a local neighborhood. The relative change of the features in the immediate surrounding of each point as well as the spatial relationships between neighboring points also need to be examined to account for spatial coherence. In this study we formulate the point labeling problem under a global graph-cut optimization solution. We construct the energy function through a graph representing a Markov Random Field (MRF). The solution to the labeling problem is obtained by finding the minimum-cut on this graph. We have employed this framework for three different labeling tasks on airborne lidar point clouds. Ground filtering, building classification, and roof-plane segmentation. As a follow-up study on our previous ground filtering work, this paper examines our building extraction approach on two airborne lidar datasets with different point densities containing approximately 930K points in one dataset and 750K points in the other. Test results for building vs. non-building point labeling show a 97.9% overall accuracy with a kappa value of 0.91 for the dataset with 1.18 pts/m2 average point density and a 96.8% accuracy with a kappa value of 0.90 for the dataset with 8.83 pts/m2 average point density. We can achieve 91.2% overall average accuracy in roof plane segmentation with respect to the reference segmentation of 20 building roofs involving 74 individual roof planes. In summary, the presented framework can successfully label points in airborne lidar point clouds with different characteristics for all three labeling problems we have introduced. It is robust to noise in the calculated features due to the use of global optimization. Furthermore, the framework achieves these results with a small training sample size.

Sensitivity analysis of parameters and contrasting performance of ground filtering algorithms with UAV photogrammetry-based and LiDAR point clouds

ABSTRACT Most ground filtering algorithms are primarily designed for airborne LiDAR point cloud processing and their successful use in identifying ground points from photogrammetric point clouds remains questionable. We compared six ground filtering algorithms implemented in Metashape, ArcGIS, CloudCompare, LAStools, and PDAL. We used UAV photogrammetry-based (acquired under leaf-off conditions) and airborne LiDAR (leaf-on) point clouds of the same area to: (i) compare accuracy of generated DTMs; (ii) evaluate the effect of vegetation density and terrain slope on filtering accuracy; and (iii) assess which algorithm parameters have the greatest effect on the filtering accuracy. Our results show that the performance of filtering algorithms was affected by the point cloud type, terrain slope and vegetation cover. The results were generally better for LiDAR (RMSE 0.13–0.19 m) than for photogrammetric (RMSE 0.19–0.23 m) point clouds. The behavior in varying vegetation and terrain conditions was consistent for LiDAR point clouds. However, when applied on photogrammetric point clouds, the algorithms’ behavior was inconsistent, especially in areas of steep slope (except for the Progressive Triangulated Irregular Network in LAStools). Parameters related to the selection of the initial minimum elevation ground points were the most influential in all algorithms and point clouds.

Positional Accuracy Assessment of Lidar Point Cloud from NAIP/3DEP Pilot Project

The Leica Geosystems CountryMapper hybrid system has the potential to collect data that satisfy the U.S. Geological Survey (USGS) National Geospatial Program (NGP) and 3D Elevation Program (3DEP) and the U.S. Department of Agriculture (USDA) National Agriculture Imagery Program (NAIP) requirements in a single collection. This research will help 3DEP determine if this sensor has the potential to meet current and future 3DEP topographic lidar collection requirements. We performed an accuracy analysis and assessment on the lidar point cloud produced from CountryMapper. The boresighting calibration and co-registration by georeferencing correction based on ground control points are assumed to be performed by the data provider. The scope of the accuracy assessment is to apply the following variety of ways to measure the accuracy of the delivered point cloud to obtain the error statistics. Intraswath uncertainty from a flat surface was computed to evaluate the point cloud precision. Intraswath difference between opposite scan directions and the interswath overlap difference were evaluated to find boresighting or any systematic errors. Absolute vertical accuracy over vegetated and non-vegetated areas were also assessed. Both horizontal and vertical absolute errors were assessed using the 3D absolute error analysis methodology of comparing conjugate points derived from geometric features. A three-plane feature makes a single unique intersection point. Intersection points were computed from ground-based lidar and airborne lidar point clouds for comparison. The difference between two intersection points form one error vector. The geometric feature-based error analysis was applied to intraswath, interswath, and absolute error analysis. The CountryMapper pilot data appear to satisfy the accuracy requirements suggested by the USGS lidar specification, based upon the error analysis results. The focus of this research was to demonstrate various conventional accuracy measures and novel 3D accuracy techniques using two different error computation methods on the CountryMapper airborne lidar point cloud.

DANCE-NET: Density-aware convolution networks with context encoding for airborne LiDAR point cloud classification

Airborne LiDAR point cloud classification has been a long-standing problem in photogrammetry and remote sensing. Early efforts either combine hand-crafted feature engineering with machine learning-based classification models or leverage the power of conventional convolutional neural networks (CNNs) on projected feature images. Recent proposed deep learning-based methods tend to develop new convolution operators which can be directly applied on raw point clouds for representative point feature learning. Although these methods have achieved satisfying performance for the classification of airborne LiDAR point clouds, they cannot adequately recognize fine-grained local structures due to the uneven density distribution of 3D point clouds. In this paper, to address this challenging issue, we introduce a density-aware convolution module which uses the point-wise density to reweight the learnable weights of convolution kernels. The proposed convolution module can approximate continuous convolution on unevenly distributed 3D point sets. Based on this convolution module, we further develop a multi-scale CNN model with downsampling and upsampling blocks to perform per-point semantic labeling. In addition, to regularize the global semantic context, we implement a context encoding module to predict a global context encoding and formulated a context encoding regularizer to enforce the predicted context encoding to be aligned with the ground truth one. The overall network can be trained in an end-to-end fashion and directly produces the desired classification results in one network forward pass. Experiments on the ISPRS 3D Labeling Dataset and 2019 Data Fusion Contest Dataset demonstrate the effectiveness and superiority of the proposed method for airborne LiDAR point cloud classification.