Using Training Samples Retrieved from a Topographic Map and Unsupervised Segmentation for the Classification of Airborne Laser Scanning Data

Classification is a fundamental task for airborne laser scanning (ALS) point cloud processing and applications. This task is challenging due to outdoor scenes with high complexity and point clouds with irregular distribution. Many existing methods based on deep learning techniques have drawbacks, such as complex pre/post-processing steps, an expensive sampling cost, and a limited receptive field size. In this paper, we propose a graph attention feature fusion network (GAFFNet) that can achieve a satisfactory classification performance by capturing wider contextual information of the ALS point cloud. Based on the graph attention mechanism, we first design a neighborhood feature fusion unit and an extended neighborhood feature fusion block, which effectively increases the receptive field for each point. On this basis, we further design a neural network based on encoder–decoder architecture to obtain the semantic features of point clouds at different levels, allowing us to achieve a more accurate classification. We evaluate the performance of our method on a publicly available ALS point cloud dataset provided by the International Society for Photogrammetry and Remote Sensing (ISPRS). The experimental results show that our method can effectively distinguish nine types of ground objects. We achieve more satisfactory results on different evaluation metrics when compared with the results obtained via other approaches.

Filtering of airborne laser scanning (ALS) point clouds into ground and nonground points is a core postprocessing step for ALS data. A hierarchical filtering method, which has high operating efficiency and accuracy because of the combination of multiscale morphology and progressive triangulated irregular network (TIN) densification (PTD), is proposed. In the proposed method, the grid is first constructed for the ALS point clouds, and virtual seed points are set by analyzing the shape and elevation distribution of points within the grid. Then, the virtual seed points are classified as ground or nonground using the multiscale morphological method. Finally, the virtual ground seed points are utilized to generate the initial TIN, and the filter is completed by iteratively densifying the initial TIN. We used various ALS data to test the performance of the proposed method. The experimental results show that the proposed filtering method has strong applicability for a variety of landscapes and, in particular, has lower commission error than the classical PTD filtering method in urban areas.

Automatic extraction of buildings from airborne laser scanning (ALS) point clouds is essential for 3D building reconstruction. This paper presents a two-part approach for extracting buildings from ALS data. First, building objects are extracted from ALS data by a marked point process using the Gibbs energy model of buildings and sampled by a reversible jump Markov chain Monte Carlo algorithm. Second, a refinement operation is performed to filter the non-building points and false building objects before extracting buildings from the detected building objects. Experimental results and evaluation using ISPRS benchmark data-sets showed the robustness of the proposed method.

Airborne laser scanning (LiDAR) point clouds over large forested areas can be processed to segment individual trees and subsequently extract tree-level information. Existing segmentation procedures typically detect more than 90% of overstory trees, yet they barely detect 60% of understory trees because of the occlusion effect of higher canopy layers. Although understory trees provide limited financial value, they are an essential component of ecosystem functioning by offering habitat for numerous wildlife species and influencing stand development. Here we model the occlusion effect in terms of point density. We estimate the fractions of points representing different canopy layers (one overstory and multiple understory) and also pinpoint the required density for reasonable tree segmentation (where accuracy plateaus). We show that at a density of ~170 pt/m² understory trees can likely be segmented as accurately as overstory trees. Given the advancements of LiDAR sensor technology, point clouds will affordably reach this required density. Using modern computational approaches for big data, the denser point clouds can efficiently be processed to ultimately allow accurate remote quantification of forest resources. The methodology can also be adopted for other similar remote sensing or advanced imaging applications such as geological subsurface modelling or biomedical tissue analysis.

<div class="section abstract"><div class="htmlview paragraph">The accident reconstruction community frequently uses Terrestrial LiDAR (TLS) to capture accurate 3D images of vehicle accident sites. This paper compares the accuracy, workflow, benefits, and challenges of Unmanned Aerial Vehicle (UAV) LiDAR, or Airborne Laser Scanning (ALS), to TLS. Two roadways with features relevant to accident reconstruction were selected for testing. ALS missions were conducted at an altitude of 175 feet and a velocity of 4 miles per hour at both sites, followed by 3D scanning using TLS. Survey control points were established to minimize error during cloud-to- cloud TLS registration and to ensure accurate alignment of ALS and TLS point clouds.</div><div class="htmlview paragraph">After data capture, the ALS point cloud was analyzed against the TLS point cloud. Approximately 80% of ALS points were within 1.8 inches of the nearest TLS point, with 64.8% at the rural site and 59.7% at the suburban site within 1.2 inches. These findings indicate that UAV-based LiDAR can achieve comparable accuracy to TLS in accident site documentation, offering potential advantages in efficiency, safety, and accessibility.</div></div>

Abstract. Point-wise classification of 3D point clouds is a challenging task in point cloud processing, whereas, in particular, its application to high-density point clouds needs special attention because a large number of point clouds affect computational efficiency negatively. Although deep learning based models have been gaining popularity in recent years and have reached state-of-the-art results in accuracy for point-wise classification, their requirements of the high number of training samples and computational resources make those models inefficient for high-density 3D point clouds. However, traditional machine learning classifiers require less training samples, so they are capable of reducing computational requirements, even considering the latest machine learning classifiers, particularly in ensemble learning of gradient boosting machines, the results can compete with deep learning models. In this study, we are studying the point-wise classification of high-density UAV LiDAR data and focusing on efficient feature extraction and a recent state-of-the-art gradient boosting machine learning classifier, LightGBM. Our proposed framework includes the following steps: at first, we are using point cloud sampling for creating sub-sampled point clouds, then we are calculating the features based on those scales implemented on GPU. Finally, we are using the LightGBM classifier for training and testing. For the evaluation of our framework, we used a publicly available benchmark dataset, Hessigheim 3D. According to the results, we achieved an overall accuracy of 87.59% and an average F1 score of 75.92%. Our framework has promising results and scores closer to deep learning models. However, more distinctive features are required to obtain more accurate results.

Deep learning in 3-D classification tasks focuses on the designs of comprehensive local aggregation operators. Airborne laser scanning (ALS) point clouds have their own characteristics: 1) the object overlaps in the vertical direction; 2) the spatial density is uneven; and 3) the objects present large-scale variations between different categories. However, the dynamic graph convolutional neural network (DGCNN) ignores the inherent properties of ALS point clouds. This study modifies the DGCNN network and proposes a new neural network module called Webconv, which densely connects point pairs in multiscale local regions to learn contextual information. One modified cross entropy loss function with variable weight is proposed to solve the problem of uneven category distributions in ALS points. Because preprocessing such as block partition ignores the context information of the whole region, the conditional random field is used to refine the point cloud. Our approach achieves state-of-the-art performance on the dataset of the 2019 IEEE GRSS Data Fusion Contest 3-D Point Cloud Classification Challenge and can be widely used for ALS point classification.

Abstract. Semantic instance segmentation from scenes, serving as a crucial role for 3D modelling and scene understanding. Conducting semantic segmentation before grouping instances is adopted by the existing state-of-the-art methods. However, without additional refinement, semantic errors will fully propagate into the grouping stage, resulting in low overlap with the ground truth instance. Furthermore, the proposed methods focused on indoor level scenes, which are limited when directly applied to large-scale outdoor Airborne Laser Scanning (ALS) point clouds. Numerous instances, significant object density and scale variations make ALS point clouds distinct from indoor data. In order to address the problems, we proposed a geometric characterization-aware semantic instance segmentation network, which utilized both semantic and objectness score to select potential points for grouping. And in point cloud feature learning stage, hand-craft geometry features are taken as input for geometric characterization awareness. Moreover, to address errors propagated from previous modules after grouping, we have additionally designed a per-instance refinement module. To assess semantic instance segmentation, we conducted experiments on an open-source dataset. Additionally, we performed semantic segmentation experiments to evaluate the performance of our proposed point cloud feature learning method.

Various methods have been developed to measure the physical presence of objects in a landscape with high positional accuracy. A new method that has been gaining popularity is Airborne Laser Scanning (ALS). ALS works by scanning a landscape (the collection of ground, buildings, vegetation, etc.,) in multiple passes. In each scan pulses of laser light are emitted from an airborne platform and their return time is measured, thus enabling the range from the point of emission to the landscape to be determined. The product of airborne laser scanning is a cloud of points in 3D space. ALS is capable of delivering very dense and accurate point clouds of a landscape in a relatively short time. However, in spite of the ability to measure objects with high positional accuracy, the automatic detection and interpretation of individual objects in landscapes remains a challenge. An example of just such a challenge is the classification of point clouds produced by ALS. The classification of ALS point clouds consists firstly in the labeling of points as either object or bare earth. The labeled object points are then further labeled as either building or vegetation. As a measurement technique ALS holds great promise and motivated by the desire to promote it, research has been conducted here to automate the detection of bare earth, buildings and vegetation in ALS point clouds.

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.

The contours, one of the most significant human perceptual features, have a significant impact on point cloud processing. In urban scenes, contour extraction is quite challenging due to the enormous number of unstructured and irregular points (typically greater than 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">7</sup> points). In this paper, we propose a Large-scale 3D point cloud Contour Extraction Network (LCE-NET) to generate contours consistent with human perception of outdoor scenes. To our knowledge, it is the first time that an end-to-end learning-based framework has been proposed for contour extraction on point cloud at this scale. The proposed LCE-Net is essentially a two-phase system. In the first phase, potential vertexes are detected from the input point cloud by the vertex detection module, then in the second phase, a designed overcomplete line proposal set is generated, and invalid line segments are further suppressed by the line proposal discrimination module. The two phases are jointly trained by a uniform loss function to promote the information interchange, thus leading to extraction results with satisfied accurate and false alarm ratings. Since there is hardly any available dataset with labeled contours for the large-scale outdoor scene, we open sourced SemanticLine, the first dataset for large-scale point clouds with labeled contour information, based on re-annotation of previous mapping level point cloud dataset semantic3D. Experimental results demonstrate that LCE-NET can effectively extract parametric contour lines from large-scale point clouds of urban scenes. Additionally, it outperforms the state-of-the-art approaches. The code will be open source on GitHub soon.

The existing deep learning methods for point cloud classification are trained using abundant labeled samples and used to test only a few samples. However, classification tasks are diverse, and not all tasks have enough labeled samples for training. In this paper, a novel point cloud classification method for indoor components using few labeled samples is proposed to solve the problem of the requirement for abundant labeled samples for training with deep learning classification methods. This method is composed of four parts: mixing samples, feature extraction, dimensionality reduction, and semantic classification. First, the few labeled point clouds are mixed with unlabeled point clouds. Next, the mixed high-dimensional features are extracted using a deep learning framework. Subsequently, a nonlinear manifold learning method is used to embed the mixed features into a low-dimensional space. Finally, the few labeled point clouds in each cluster are identified, and semantic labels are provided for unlabeled point clouds in the same cluster by a neighborhood search strategy. The validity and versatility of the proposed method were validated by different experiments and compared with three state-of-the-art deep learning methods. Our method uses fewer than 30 labeled point clouds to achieve an accuracy that is 1.89–19.67% greater than existing methods. More importantly, the experimental results suggest that this method is not only suitable for single-attribute indoor scenarios but also for comprehensive complex indoor scenarios.

Reliable information concerning stand volume is fundamental to making strategic decisions in sustainable forest management. A variety of remotely sensed data and different inventory methods have been used for the estimation of forest biometric parameters. Particularly, airborne laser scanning (ALS) point clouds are widely used for the estimation of stand volume and forest biomass using an area-based approach (ABA) framework. This method relies on the reference measurements of field plots with the necessary prerequisite of a precise co-registration between ground reference plots and the corresponding ALS samples. In this research, the allometric area-based approach (AABA) is proposed in the context of stand volume estimation of Scots pine (Pinus sylvestris L.) stands. The proposed method does not require detailed information about the coordinates of the field plots. We applied Polish National Forest Inventory data from 9400 circular field plots (400 m2) to develop a plot level stand volume allometric model using two independent variables: top height (TH) and relative spacing index (RSI). The model was developed using the multiple linear regression method with a log–log transformation of variables. The hypothesis was that, the field measurements of TH and RSI could be replaced with corresponding ALS-derived metrics. It was assumed that TH could be represented by the maximum height of the ALS point cloud, while RSI can be calculated based on the number of tree crowns delineated within the ALS-derived canopy height model. Performance of the developed AABA model was compared with the semi-empirical ABASE (with two predictors: TH and RSI) and empirical ABAE (several point cloud metrics as predictors). The models were validated at the plot level using 315 forest management inventory plots (400 m2) and at the stand level using the complete field measurements from 42 Scots pine dominated forest stands in the Milicz forest district (Poland). The AABA model showed a comparable accuracy to the traditional ABA models with relatively high accuracy at the plot (relative root mean square error (RMSE) = 22.8 per cent; R2 = 0.63) and stand levels (RMSE = 17.8 per cent, R2 = 0.65). The proposed novel approach reduces time- and cost-consuming field work required for the classic ABA method, without a significant reduction in the accuracy of stand volume estimations. The AABA is potentially applicable in the context of forest management inventory without the necessity for field measurements at local scale. The transportability of the approach to other species and more complex stands needs to be explored in future studies.

Airborne laser scanning (ALS) and digital aerial photogrammetry (DAP) have both been demonstrated as reliable sources of information on forest stand inventory attributes. The increasing availability of both datasets provides a means for improving stand dynamics information over time; however, the cost of multi-temporal ALS can be prohibitive in some circumstances. As a result, a combination of ALS at an initial time step and subsequent updates using DAP has been proposed as a cost-effective alternative for maintaining forest inventories. In this study we used low density ALS and DAP point clouds acquired in 2007 and 2015, respectively, to quantify changes in forest structure, in a highly disturbed boreal mixedwood forest in Alberta, Canada. We examined the capacity of the two technologies to model changes in top height (H), volume (V), and basal area (BA) using both direct and indirect approaches for estimation. Results indicate that the proportion of explained variance (adjusted R2) for the models derived from the ALS (Time 1; T1) and DAP (Time 2; T2) data were highest for models predicting H at T1, and lowest for BA at T1 and T2 (R2 was 0.66–0.70). The indirect estimates of change in H, BA, and V were calculated by subtracting the T1 and T2 predictions. For the direct approach, separate regression models were developed that used the differences in point cloud metrics between T1 and T2 as predictors. Results indicated that the accuracy of the estimates generated using the indirect approach were markedly lower than the estimates generated using the direct approach, with especially poor results for ∆BA and ∆V. Best results were achieved for ∆H using the direct approach with an R2 coefficient of 0.65 and an root mean square error (RMSE)% of 190.06%. We found that the error associated with change estimates of H, BA and V increased with the increase or decrease in mortality. We conclude that forest managers should act carefully when applying the multi-temporal and multi-sensor analysis of forest growth if forest growth is slow and the level of mortality is high.

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