A simple oriented search and clustering method for extracting individual forest trees from ALS point clouds

Power lines classification is important for electric power management and geographical objects extraction using LiDAR (light detection and ranging) point cloud data. Many supervised classification approaches have been introduced for the extraction of features such as ground, trees, and buildings, and several studies have been conducted to evaluate the framework and performance of such supervised classification methods in power lines applications. However, these studies did not systematically investigate all of the relevant factors affecting the classification results, including the segmentation scale, feature selection, classifier variety, and scene complexity. In this study, we examined these factors systematically using airborne laser scanning and mobile laser scanning point cloud data. Our results indicated that random forest and neural network were highly suitable for power lines classification in forest, suburban, and urban areas in terms of the precision, recall, and quality rates of the classification results. In contrast to some previous studies, random forest yielded the best results, while Naïve Bayes was the worst classifier in most cases. Random forest was the more robust classifier with or without feature selection for various LiDAR point cloud data. Furthermore, the classification accuracies were directly related to the selection of the local neighborhood, classifier, and feature set. Finally, it was suggested that random forest should be considered in most cases for power line classification.

Learning a constrained conditional random field for enhanced segmentation of fallen trees in ALS point clouds

<strong class="journal-contentHeaderColor">Abstract.</strong> This study investigates rock glacier destabilization based on the results of a unique in situ and remote-sensing-based monitoring network focused on the kinematics of the rock glacier in Äußeres Hochebenkar (Austrian Alps). We consolidate, homogenize, and extend existing time series to generate a comprehensive dataset consisting of 14 digital surface models covering a 68-year time period, as well as in situ measurements of block displacement since the early 1950s. The digital surface models are derived from historical aerial imagery and, more recently, airborne and uncrewed-aerial-vehicle-based laser scanning (ALS and ULS, respectively). High-resolution 3D ALS and ULS point clouds are available at annual temporal resolution from 2017 to 2021. Additional terrestrial laser scanning data collected in bi-weekly intervals during the summer of 2019 are available from the rock glacier front. Using image correlation techniques, we derive velocity vectors from the digital surface models, thereby adding rock-glacier-wide spatial context to the point-scale block displacement measurements. Based on velocities, surface elevation changes, analyses of morphological features, and computations of the bulk creep factor and strain rates, we assess the combined datasets in terms of rock glacier destabilization. To additionally investigate potential rotational components of the movement of the destabilized section of the rock glacier, we integrate in situ data of block displacement with ULS point clouds and compute changes in the rotation angles of single blocks during recent years. The time series shows two cycles of destabilization in the lower section of the rock glacier. The first lasted from the early 1950s until the mid-1970s. The second began around 2017 after approximately 2 decades of more gradual acceleration and is currently ongoing. Both destabilization periods are characterized by high velocities and the development of morphological destabilization features on the rock glacier surface. Acceleration in the most recent years has been very pronounced, with velocities reaching 20–30 m a<span class="inline-formula">^−1</span> in 2020–2021. These values are unprecedented in the time series and suggest highly destabilized conditions in the lower section of the rock glacier, which shows signs of translational and rotational landslide-like movement. Due to the length and granularity of the time series, the cyclic destabilization process at the Äußeres Hochebenkar rock glacier is well resolved in the dataset. Our study highlights the importance of interdisciplinary, long-term, and continuous high-resolution 3D monitoring to improve process understanding and model development related to rock glacier rheology and destabilization.

Individual tree detection from airborne laser scanning data based on supervoxels and local convexity

Detailed forest inventory and mensuration of individual trees have drawn attention of research society mainly to support sustainable forest management. This study aims at estimating individual tree attributes from high density point cloud obtained by terrestrial laser scanner (TLS). The point clouds were obtained over single reference tree and group of trees in forest area. The reference tree is treated as benchmark since detailed measurements of branch diameter were made on selected branches with different sizes and locations. Diameter at breast height (DBH) was measured for trees in forest. Furthermore tree height, height to crown base, crown volume and tree branch volume were also estimated for each tree. Branch diameter is estimated directly from the point clouds based on semi-automatic approach of model fitting i.e. sphere, ellipse and cylinder. Tree branch volume is estimated based on the volume of the fitted models. Tree height and height to crown base are computed using histogram analysis of the point clouds elevation. Tree crown volume is estimated by fitting a convex-hull on the tree crown. The results show that the Root Mean Squared Error (RMSE) of the estimated tree branch diameter does not have a specific trend with branch sizes and number of points used for fitting process. This explains complicated distribution of point clouds over the branches. Overall cylinder model produces good results with most branch sizes and number of point clouds for fitting. The cylinder fitting approach shows significantly better estimation results compared to sphere and ellipse fitting models.

Information about trees in forest is essential for the assessment of the quantity and the quality of forest ecosystem services. Recently, Deep Learning (DL) methods were regarded as a new cornerstone of algorithm development. Semantic annotations of 3D forest scenes are fundamental for DL algorithm developments. Its necessity has become more urgent as DL is data-driven and requires large amount of training and verification data. However, high-quality annotated forest datasets are still rare, as trees comprise of irregular structures and small components and pose significantly greater challenges even for manual recognition in comparison with artificial objects. This paper introduces a new open point cloud dataset ForestSemantic for forest semantic studies at both individual tree- and plot-levels. The dataset is based on TLS data with different forest conditions. Manual annotation was carried out to a level of detail 4, i.e., until all visible branches. Semantic information is provided at both plot- and tree-levels, as well as at both object- and point-level. Thus, the dataset supports both instance and semantic studies, such as objects detection and segmentation and classification at both tree- and plot-levels. In addition, the dataset also provides comprehensive structural tree traits as reference for further methodological development and verification. This dataset is expected to facilitate research in new dimensions and benchmarks of different systems and solutions. A few examples are demonstrated in this paper to unveil the potentials of the dataset for various applications. In future, it is also possible to simulate other types of point clouds by down-sampling and deforming, and to transfer the dataset for training and verification of other close-range sensing systems, as the dataset was generated using TLS point clouds that represent the highest spatial resolution and geometric accuracy in all close-range point clouds.

Segmentation of vegetation LiDAR point clouds is an important method for obtaining individual tree structure parameters. The current individual tree segmentation methods are mainly for airborne LiDAR point clouds, which use elevation information to form a grid map for segmentation, or use canopy vertices as seed points for clustering. Side-view LiDAR (vehicle LiDAR and hand-held LiDAR) can acquire more information about the lower layer of trees, but it is a challenge to perform the individual tree segmentation because the structure of side-view LiDAR point clouds is more complex. This paper proposes an individual tree segmentation method called Shadow-cut to extract the contours of the street tree point cloud. Firstly, we separated the region of the trees using the binary classifier (e.g., support vector machine) based on point cloud geometric features. Then, the optimal projection of the 3D point clouds to the 2D image is calculated and the optimal projection is the case where the pixels of the individual tree image overlap the least. Finally, after using the image segmentation algorithm to extract the tree edges in the 2D image, the corresponding 3D individual tree point cloud contours are matched with the pixels of individual tree edges in the 2D image. We conducted experiments with the proposed method on LiDAR data of urban street trees, and the correctness, completeness, and quality of the proposed individual tree segmentation method reached 91.67%, 85.33%, and 79.19%, which were superior to the CHM-based method by 2.70%, 6.19%, and 7.12%, respectively. The results show that this method is a practical and effective solution for individual tree segmentation in the LiDAR point clouds of street trees.

Individual tree (IT) segmentation is crucial for forest management, supporting forest inventory, biomass monitoring or tree competition analysis. Light detection and ranging (LiDAR) is a prominent technology in this context, outperforming competing technologies. Aerial laser scanning (ALS) is frequently used for forest documentation, showing good point densities at the tree-top surface. Even though under-canopy data collection is possible with multi-echo ALS, the number of points for regions near the ground in leafy forests drops drastically, and, as a result, terrestrial laser scanners (TLS) may be required to obtain reliable information about tree trunks or under-growth features. In this work, an IT extraction method for terrestrial backpack LiDAR data is presented. The method is based on DBSCAN clustering and cylinder voxelization of the volume, showing a high detection rate (∼90%) for tree locations obtained from point clouds, and low commission and submission errors (accuracy over ). The method includes a sensibility assessment to calculate the optimal input parameters and adapt the workflow to real-world data. This approach shows that forest management can benefit from IT segmentation, using a handheld TLS to improve data collection productivity.

Light Detection and Ranging (LiDAR) has advantages in detecting individual trees because it can obtain information on the vertical structure and even on the lower layers. However, the current methods still cannot detect understory well, because the small trees are often clumped together and overlapped by large trees. To fill this gap, a two-stage network named Tree Region-Based Convolutional Neural Network (RCNN) was proposed to directly detect trees from point clouds. In the first stage, very dense anchors are generated anywhere in a forest. Then, Tree RCNN can directly focus on determining whether an anchor belongs to an individual tree or not and generate tree proposals based on the anchors. In this way, the small trees overlapped by big trees can be detected in the process. In the second stage, multi-position feature extraction is proposed to extract shape features of the tree proposals output in the first stage to refine the tree proposals. The positions and heights of detected trees can be obtained by the refined tree proposals. The performance of our method was estimated by a public dataset. Compared to methods provided by the dataset and the commonly used deep learning methods, Tree RCNN achieved the best performance, especially for the lower-layer trees. The root mean square value of detection rates (RMSass) of all plots of the dataset reached 61%, which was 6 percentage points higher than the best RMSass of other methods. The RMSass of the layers &lt; 5 m, 5–10 m, 10–15 m, and 15–20 reached 20%, 38%, 48%, and 61%, which was 5, 6, 7, and 3 percentage points higher than the best RMSass of other methods, respectively. The results indicated our method can be a useful tool for tree detection.

Hurricane Maria crossed Puerto Rico with winds as high as 250 km/h, resulting in widespread damages and loss of weather station data, thus limiting direct weather measurements of wind variability. Here, we identified more than 155 million trees to estimate the distribution of wind speed over 9000 km2 of land from island-wide LiDAR point clouds collected before and after the hurricane. The point clouds were classified and rasterized into the canopy height model to perform individual tree identification and perform change detection analysis. Individual trees’ stem diameter at breast height were estimated using a function between delineated crown and extracted canopy height, validated using the records from Puerto Rico’s Forest Inventory 2003. The results indicate that approximately 35.7% of trees broke at the stem (below the canopy center) and 28.5% above the canopy center. Furthermore, we back-calculated the critical wind speed, or the minimum speed to cause breakage, at individual tree level this was performed by applying a mechanical model using the estimated diameter at breast height, the extrapolated breakage height, and pre-Hurricane Maria canopy height. Individual trees were then aggregated at 115 km2 cells to summarize the critical wind speed distribution of each cell, based on the percentage of stem breakage. A vertical wind profile analysis was then applied to derive the hurricane wind distribution using the mean hourly wind speed 10 m above the canopy center. The estimated wind speed ranges from 250 km/h in the southeast at the landfall to 100 km/h in the southwest parts of the islands. Comparison of the modeled wind speed with the wind gust readings at the few remaining NOAA stations support the use of tree breakages to model the distribution of hurricane wind speed when ground readings are sparse.

Trees are an integral component of the urban environment and important for human well-being, adaption measures to climate change and sustainable urban transformation. Understanding the small-scale impacts of urban trees and strategically managing the ecosystem services they provide requires high-resolution information on urban forest structure, which is still scarce. In contrast, there is an abundance of data portraying urban areas and an associated trend towards smart cities and digital twins as analysis platforms. A GIS workflow is presented in this paper that may close this data gap by classifying the urban forest from LiDAR point clouds, detecting and reconstructing individual crowns, and enabling a tree representation within semantic 3D city models. The workflow is designed to provide robust results for point clouds with a density of at least 4 pts/m2 that are widely available. Evaluation was conducted by mapping the urban forest of Dresden (Germany) using a point cloud with 4 pts/m². An object-based data fusion approach is implemented for the classification of the urban forest. A classification accuracy of 95 % for different urban settings is achieved by combining LiDAR with multispectral imagery and a 3D building model. Individual trees are detected by local maxima filtering and crowns are segmented using marker-controlled watershed segmentation. Evaluation highlights the influences of both urban and forest structure on individual tree detection. Substantial differences in detection accuracies are evident between trees along streets (72 %) and structurally more complex tree stands in green areas (31 %), as well as dependencies on tree height and crown diameter. Furthermore, an approach for parameterized reconstruction of tree crowns is presented, which enables efficient and realistic city-wide modeling. The suitability of LiDAR to measure individual tree metrics is illustrated as well as a framework for modeling individual tree crowns via geometric primitives.

The individual tree information is the most important parameter of biomass inversion. Recently, LiDAR has been widely and successfully applied in forest research, and it shows promise to map individual trees in complex and heterogeneous forests. Based on Airborne LiDAR point cloud, this paper uses local maximum filtering technique to extract the height and crown of individual tree of Qinghai spruce forest in Qilian Mountains. The analysis shows that accuracy of the three methods is depended on the Thresholds. All of them only detect less than 50% trees in the study area. There are two reasons. The first is the density of the point clouds is only 6 dot/m2. The second is some trees are understory and the height is too small, even less than the error range of CHM.

Tree localization in point clouds of forest scenes is critical in the forest inventory. Most of the existing methods proposed for TLS forest data are based on model fitting or point-wise features which are time-consuming, sensitive to data incompleteness and complex tree structures. Furthermore, these methods often require lots of preprocessing such as ground filtering and noise removal. The fast and easy-to-use top-based methods that are widely applied in processing ALS point clouds are not applicable in localizing trees in TLS point clouds due to the data incompleteness and complex canopy structures. The objective of this study is to make the top-based methods applicable to TLS forest point clouds. To this end, a novel point cloud transformation is presented, which enhances the visual salience of tree instances and makes the top-based methods adapting to TLS forest scenes. The input for the proposed method is the raw point clouds and no other pre-processing steps are needed. The new method is tested on an international benchmark and the experimental results demonstrate its necessity and effectiveness. Finally, the proposed method has the potential to benefit other object localization tasks in different scenes based on detailed analysis and tests.

Over an extended period, remote-sensing-based individual tree analysis has played a critical role in modern forest inventory and management research. The segmentation of individual trees from aerial point clouds usually depends on the characteristics of peak-like uplift on the crown surface; however, the performance inevitably decreases with increasing visibility of such features in point clouds, especially for high-density forests. Herein, we developed a novel hierarchical region-merging algorithm that first over-segmented the entire forest scene based on local density and then merged the over-segmented partitions into pairs through a stepwise optimal process to produce the final segmentation. In the region-merging method, a global merging cost was introduced to shift from local detection of crown features to use the overall compactness of forest point clouds. The experiments were conducted using unmanned aerial vehicle light detection and ranging (UAV-LiDAR) point clouds from three coniferous stands with different densities and a high-density coniferous and broad-leaved mixed stand. A total of 5510 field-measured trees in 36 plots were used to assess the accuracy of the proposed method. Our method achieved F-scores of 0.91, 0.88, 0.84, and 0.80 for low- (~700 stems/ha), medium- (~1000 stems/ha), and high-density (~2000 stems/ha) conifer stands and coniferous and broad-leaved mixed forests (~1800 stems/ha), respectively. Compared with the classical individual tree segmentation methods (marker-controlled watershed segmentation and point cloud region-growing algorithm), our method obtained comparable performance in low-density conifer stands and superior performance in the other stands. Furthermore, the region-merging algorithm could detect 10% more suppressed trees on average, which led to an apparent improvement in detection accuracy. The proposed algorithm provides a flexible segmentation framework that could be further improved by a different design that merges costs or applies multiscale segmentation with different stopping criteria.

To compare the capability and the accuracy of Light Detection And Ranging (lidar) and Interferometric Synthetic Aperture Radar (INSAR) for the detection and measurement of individual tree heights and forest plot heights, one lidar dataset with nominal spacing of 3 m and one short-wavelength Ku-band INSAR with comparable ground resolution of 3 m were studied. Vegetation heights were based on the subtraction between the bare ground Digital Elevation Models (DEMs) and the canopy Digital Surface Models (DSMs). Two field measurement datasets of isolated individual trees and forest plots were used for validation. Results showed 78 percent and 17 percent of the individual trees was detected on lidar and INSAR data, respectively. General canopy height patterns could be successfully identified through a transect profile for both sensors, however, both sensors consistently underestimated vegetation height. Higher accuracy was obtained for both individual tree level and forest plot level with lidar than that from the Ku-band INSAR. These results indicate that lidar is much better than INSAR for the detection and estimation of tree and forest plot height.