scholarly journals Predicting Tree Species From 3D Laser Scanning Point Clouds Using Deep Learning

2021 ◽  
Vol 12 ◽  
Author(s):  
Dominik Seidel ◽  
Peter Annighöfer ◽  
Anton Thielman ◽  
Quentin Edward Seifert ◽  
Jan-Henrik Thauer ◽  
...  
Keyword(s):  
Tree Species ◽  
Classification Accuracy ◽  
Laser Scanning ◽  
Confusion Matrix ◽  
Point Clouds ◽  
Small Sample ◽  
Training Data ◽  
Computationally Efficient ◽  
Species Classification ◽  
3D Point Clouds

Automated species classification from 3D point clouds is still a challenge. It is, however, an important task for laser scanning-based forest inventory, ecosystem models, and to support forest management. Here, we tested the performance of an image classification approach based on convolutional neural networks (CNNs) with the aim to classify 3D point clouds of seven tree species based on 2D representation in a computationally efficient way. We were particularly interested in how the approach would perform with artificially increased training data size based on image augmentation techniques. Our approach yielded a high classification accuracy (86%) and the confusion matrix revealed that despite rather small sample sizes of the training data for some tree species, classification accuracy was high. We could partly relate this to the successful application of the image augmentation technique, improving our result by 6% in total and 13, 14, and 24% for ash, oak and pine, respectively. The introduced approach is hence not only applicable to small-sized datasets, it is also computationally effective since it relies on 2D instead of 3D data to be processed in the CNN. Our approach was faster and more accurate when compared to the point cloud-based “PointNet” approach.

scholarly journals VALIDATING A WORKFLOW FOR TREE INVENTORY UPDATING WITH 3D POINT CLOUDS OBTAINED BY MOBILE LASER SCANNING

2018 ◽  
Vol XLII-2 ◽  
pp. 1163-1168
Author(s):  
J. Wang ◽  
R. Lindenbergh
Keyword(s):  
Tree Species ◽  
Laser Scanning ◽  
Spatial Information ◽  
Open Data ◽  
Point Clouds ◽  
Urban Trees ◽  
Biomass Estimation ◽  
Species Classification ◽  
3D Point Clouds ◽  
Tree Inventory

Urban trees are an important component of our environment and ecosystem. Trees are able to combat climate change, clean the air and cool the streets and city. Tree inventory and monitoring are of great interest for biomass estimation and change monitoring. Conventionally, parameters of trees are manually measured and documented in situ, which is not efficient regarding labour and costs. Light Detection And Ranging (LiDAR) has become a well-established surveying technique for the acquisition of geo-spatial information. Combined with automatic point cloud processing techniques, this in principle enables the efficient extraction of geometric tree parameters. In recent years, studies have investigated to what extend it is possible to perform tree inventories using laser scanning point clouds. Give the availability of a city of Delft Open data tree repository, we are now able to present, validate and extend a workflow to automatically obtain tree data from tree location until tree species. The results of a test over 47 trees show that the proposed methods in the workflow are able to individual urban trees. The tree species classification results based on the extracted tree parameters show that only one tree was wrongly classified using k-means clustering.

scholarly journals FEATURE RELEVANCE ASSESSMENT OF MULTISPECTRAL AIRBORNE LIDAR DATA FOR TREE SPECIES CLASSIFICATION

2018 ◽  
Vol XLII-3 ◽  
pp. 31-34 ◽  
Author(s):  
N. Amiri ◽  
M. Heurich ◽  
P. Krzystek ◽  
A. K. Skidmore
Keyword(s):  
Tree Species ◽  
Laser Scanning ◽  
Point Clouds ◽  
Airborne Lidar ◽  
Species Classification ◽  
Tree Species Classification ◽  
3D Point Clouds ◽  
Segmentation Approach ◽  
Species Mapping ◽  
Spectral Channels

The presented experiment investigates the potential of Multispectral Laser Scanning (MLS) point clouds for single tree species classification. The basic idea is to simulate a MLS sensor by combining two different Lidar sensors providing three different wavelngthes. The available data were acquired in the summer 2016 at the same date in a leaf-on condition with an average point density of 37&amp;thinsp;points/m<sup>2</sup>. For the purpose of classification, we segmented the combined 3D point clouds consisiting of three different spectral channels into 3D clusters using Normalized Cut segmentation approach. Then, we extracted four group of features from the 3D point cloud space. Once a varity of features has been extracted, we applied forward stepwise feature selection in order to reduce the number of irrelevant or redundant features. For the classification, we used multinomial logestic regression with <i>L<sub>1</sub></i> regularization. Our study is conducted using 586 ground measured single trees from 20 sample plots in the Bavarian Forest National Park, in Germany. Due to lack of reference data for some rare species, we focused on four classes of species. The results show an improvement between 4&amp;ndash;10&amp;thinsp;pp for the tree species classification by using MLS data in comparison to a single wavelength based approach. A cross validated (15-fold) accuracy of 0.75 can be achieved when all feature sets from three different spectral channels are used. Our results cleary indicates that the use of MLS point clouds has great potential to improve detailed forest species mapping.

scholarly journals SEMANTIC LABELING OF ALS POINT CLOUDS FOR TREE SPECIES MAPPING USING THE DEEP NEURAL NETWORK POINTNET++

2019 ◽  
Vol XLII-2/W13 ◽  
pp. 951-955 ◽  
Author(s):  
S. Briechle ◽  
P. Krzystek ◽  
G. Vosselman
Keyword(s):  
Tree Species ◽  
Laser Scanning ◽  
Large Scale ◽  
Point Clouds ◽  
Training Data ◽  
Deciduous Trees ◽  
Individual Tree ◽  
Coniferous Trees ◽  
3D Point Clouds ◽  
Semantic Labeling

<p><strong>Abstract.</strong> Most methods for the mapping of tree species are based on the segmentation of single trees that are subsequently classified using a set of hand-crafted features and an appropriate classifier. The classification accuracy for coniferous and deciduous trees just using airborne laser scanning (ALS) data is only around 90% in case the geometric information of the point cloud is used. As deep neural networks (DNNs) have the ability to adaptively learn features from the underlying data, they have outperformed classic machine learning (ML) approaches on well-known benchmark datasets provided by the robotics, computer vision and remote sensing community. Though, tree species classification using deep learning (DL) procedures has been of minor research interest so far. Some studies have been conducted based on an extensive prior generation of images or voxels from the 3D raw data. Since innovative DNNs directly operate on irregular and unordered 3D point clouds on a large scale, the objective of this study is to exemplarily use PointNet++ for the semantic labeling of ALS point clouds to map deciduous and coniferous trees. The dataset for our experiments consists of ALS data from the Bavarian Forest National Park (366 trees/ha), only including spruces (coniferous) and beeches (deciduous). First, the training data were generated automatically using a classic feature-based Random Forest (RF) approach classifying coniferous trees (precision&amp;thinsp;=&amp;thinsp;93%, recall&amp;thinsp;=&amp;thinsp;80%) and deciduous trees (precision&amp;thinsp;=&amp;thinsp;82%, recall&amp;thinsp;=&amp;thinsp;92%). Second, PointNet++ was trained and subsequently evaluated using 80 randomly chosen test batches à 400&amp;thinsp;m<sup>2</sup>. The achieved per-point classification results after 163 training epochs for coniferous trees (precision&amp;thinsp;=&amp;thinsp;90%, recall&amp;thinsp;=&amp;thinsp;79%) and deciduous trees (precision&amp;thinsp;=&amp;thinsp;81%, recall&amp;thinsp;=&amp;thinsp;91%) are fairly high considering that only the geometry was included. Nevertheless, the classification results using PointNet++ are slightly lower than those of the baseline method using a RF classifier. Errors in the training data and occurring edge effects limited a better performance. Our first results demonstrate that the architecture of the 3D DNN PointNet++ can successfully be adapted to the semantic labeling of large ALS point clouds to map deciduous and coniferous trees. Future work will focus on the integration of additional features like i.e. the laser intensity, the surface normals and multispectral features into the DNN. Thus, a further improvement of the accuracy of the proposed approach is to be expected. Furthermore, the classification of numerous individual tree species based on pre-segmented single trees should be investigated.</p>

scholarly journals EVALUATING THE POSSIBILITY OF TREE SPECIES CLASSIFICATION WITH DUAL-WAVELENGTH ALS DATA

2019 ◽  
Vol XLII-2/W13 ◽  
pp. 1097-1103
Author(s):  
M. Pilarska ◽  
W. Ostrowski
Keyword(s):  
Spatial Data ◽  
Tree Species ◽  
Laser Scanning ◽  
Laser Scanner ◽  
Point Clouds ◽  
Airborne Laser Scanning ◽  
Tree Canopy ◽  
Species Classification ◽  
Airborne Laser ◽  
Simultaneous Acquisition

<p><strong>Abstract.</strong> Airborne laser scanning (ALS) plays an important role in spatial data acquisition. One of the advantages of this technique is laser beam penetration through vegetation, which makes it possible to not only obtain data on the tree canopy but also within and under the canopy. In recent years, multi-wavelength airborne laser scanning has been developed. This technique consists of simultaneous acquisition of point clouds in more than one band. The aim of this experiment was to examine and assess the possibilities of tree segmentation and species classification in an urban area. In this experiment, point clouds registered in two wavelengths (532 and 1064&amp;thinsp;nm) were used for tree segmentation and species classification. The data were acquired with a Riegl VQ-1560i-DW laser scanner over Elblag, Poland, during August 2018. Tree species collected by a botanist team within terrain measurements were used as a reference in the classification process. Within the experiment segmentation and classification process were performed. Regarding the segmentation, TerraScan software and Li et al.’s algorithm, implemented in LidR package were used. Results from both methods are clearly over-segmented in comparison to the manual segments. In Terrasolid segmentation, single reference segments are over-segmented in 28% of cases, whereas, for LidR, over-segmentation occurred in 73% of the segments. According the classification results, Thuja, Salix and Betula were the species, for which the highest classification accuracy was achieved.</p>

Tropical Tree Species 3D Modelling and Classification Based on LiDAR Technology

2020 ◽  
pp. 1-22
Author(s):  
Panagiotis Barmpoutis ◽  
Tania Stathaki ◽  
Jonathan Lloyd ◽  
Magna Soelma Bessera de Moura
Keyword(s):  
Tree Species ◽  
Laser Scanning ◽  
Tensor Decomposition ◽  
Singular Values ◽  
Point Clouds ◽  
Tropical Tree ◽  
Species Classification ◽  
Tropical Tree Species ◽  
Sensing Technology ◽  
3D Information

Over the last decade or so, laser scanning technology has become an increasingly popular and important tool for forestry inventory, enabling accurate capture of 3D information in a fast and environmentally friendly manner. To this end, the authors propose here a system for tropical tree species classification based on 3D scans of LiDAR sensing technology. In order to exploit the interrelated patterns of trees, skeleton representations of tree point clouds are extracted, and their structures are divided into overlapping equal-sized 3D segments. Subsequently, they represent them as third-order sparse structure tensors setting the value of skeleton coordinates equal to one. Based on the higher-order tensor decomposition of each sparse segment, they 1) estimate the mode-n singular values extracting intra-correlations of tree branches and 2) model tropical trees as linear dynamical systems extracting appearance information and dynamics. The proposed methodology was evaluated in tropical tree species and specifically in a dataset consisting of 26 point clouds of common Caatinga dry-forest trees.

scholarly journals Single Tree Classification Using Multi-Temporal ALS Data and CIR Imagery in Mixed Old-Growth Forest in Poland

2021 ◽  
Vol 13 (24) ◽  
pp. 5101
Author(s):  
Agnieszka Kamińska ◽  
Maciej Lisiewicz ◽  
Krzysztof Stereńczak
Keyword(s):  
Tree Species ◽  
Classification Accuracy ◽  
Sustainable Forest Management ◽  
Point Clouds ◽  
Deciduous Tree ◽  
Spectral Information ◽  
Species Classification ◽  
Individual Tree ◽  
Mean Values ◽  
Old Growth Forest

Tree species classification is important for a variety of environmental applications, including biodiversity monitoring, wildfire risk assessment, ecosystem services assessment, and sustainable forest management. In this study we used a fusion of three remote sensing (RM) datasets including ALS (leaf-on and leaf-off) and colour-infrared (CIR) imagery (leaf-on), to classify different coniferous and deciduous tree species, including dead class, in a mixed temperate forest in Poland. We used intensity and structural variables from the ALS data and spectral information derived from aerial imagery for the classification procedure. Additionally, we tested the differences in classification accuracy of all the variants included in the data integration. The random forest classifier was used in the study. The highest accuracies were obtained for classification based on both point clouds and including image spectral information. The mean values for overall accuracy and kappa were 84.3% and 0.82, respectively. Analysis of the leaf-on and leaf-off alone is not sufficient to identify individual tree species due to their different discriminatory power. Leaf-on and leaf-off ALS point cloud features alone gave the lowest accuracies of 72% ≤ OA ≤ 74% and 0.67 ≤ κ ≤ 0.70. Classification based on both point clouds was found to give satisfactory and comparable results to classification based on combined information from all three sources (83% ≤ OA ≤ 84% and 0.81 ≤ κ ≤ 0.82). The classification accuracy varied between species. The classification results for coniferous trees were always better than for deciduous trees independent of the datasets. In the classification based on both point clouds (leaf-on and leaf-off), the intensity features seemed to be more important than the other groups of variables, especially the coefficient of variation, skewness, and percentiles. The NDVI was the most important CIR-based feature.

scholarly journals Hyperspectral tree crown classification using the multiple instance adaptive cosine estimator

PeerJ ◽  
2019 ◽  
Vol 7 ◽  
pp. e6405
Author(s):  
Sheng Zou ◽  
Paul Gader ◽  
Alina Zare
Keyword(s):  
Tree Species ◽  
Classification Accuracy ◽  
Ground Truth ◽  
Majority Voting ◽  
Training Data ◽  
Cross Entropy ◽  
Tree Crown ◽  
Species Classification ◽  
Testing Data ◽  
Ace Algorithm

Tree species classification using hyperspectral imagery is a challenging task due to the high spectral similarity between species and large intra-species variability. This paper proposes a solution using the Multiple Instance Adaptive Cosine Estimator (MI-ACE) algorithm. MI-ACE estimates a discriminative target signature to differentiate between a pair of tree species while accounting for label uncertainty. Multi-class species classification is achieved by training a set of one-vs-one MI-ACE classifiers corresponding to the classification between each pair of tree species and a majority voting on the classification results from all classifiers. Additionally, the performance of MI-ACE does not rely on parameter settings that require tuning resulting in a method that is easy to use in application. Results presented are using training and testing data provided by a data analysis competition aimed at encouraging the development of methods for extracting ecological information through remote sensing obtained through participation in the competition. The experimental results using one-vs-one MI-ACE technique composed of a hierarchical classification, where a tree crown is first classified to one of the genus classes and one of the species classes. The species-level rank-1 classification accuracy is 86.4% and cross entropy is 0.9395 on the testing data, provided by the competition organizer, without the release of ground truth for testing data. Similarly, the same evaluation metrics are computed on the training data, where the rank-1 classification accuracy is 95.62% and the cross entropy is 0.2649. The results show that the presented approach can not only classify the majority species classes, but also classify the rare species classes.

scholarly journals TREE SPECIES CLASSIFICATION OF BROADLEAVED FORESTS IN NAGANO, CENTRAL JAPAN, USING AIRBORNE LASER DATA AND MULTISPECTRAL IMAGES

2017 ◽  
Vol XLII-3/W3 ◽  
pp. 33-38
Author(s):  
S. Deng ◽  
M. Katoh ◽  
Y. Takenaka ◽  
K. Cheung ◽  
A. Ishii ◽  
...  
Keyword(s):  
Tree Species ◽  
Laser Scanning ◽  
Point Clouds ◽  
Larix Kaempferi ◽  
Multispectral Images ◽  
Central Japan ◽  
Tree Crowns ◽  
The Third ◽  
3D Point Clouds ◽  
Airborne Laser

This study attempted to classify three coniferous and ten broadleaved tree species by combining airborne laser scanning (ALS) data and multispectral images. The study area, located in Nagano, central Japan, is within the broadleaved forests of the Afan Woodland area. A total of 235 trees were surveyed in 2016, and we recorded the species, DBH, and tree height. The geographical position of each tree was collected using a Global Navigation Satellite System (GNSS) device. Tree crowns were manually detected using GNSS position data, field photographs, true-color orthoimages with three bands (red-green-blue, RGB), 3D point clouds, and a canopy height model derived from ALS data. Then a total of 69 features, including 27 image-based and 42 point-based features, were extracted from the RGB images and the ALS data to classify tree species. Finally, the detected tree crowns were classified into two classes for the first level (coniferous and broadleaved trees), four classes for the second level (<i>Pinus densiflora</i>, <i>Larix kaempferi</i>, <i>Cryptomeria japonica</i>, and broadleaved trees), and 13 classes for the third level (three coniferous and ten broadleaved species), using the 27 image-based features, 42 point-based features, all 69 features, and the best combination of features identified using a neighborhood component analysis algorithm, respectively. The overall classification accuracies reached 90&amp;thinsp;% at the first and second levels but less than 60&amp;thinsp;% at the third level. The classifications using the best combinations of features had higher accuracies than those using the image-based and point-based features and the combination of all of the 69 features.

scholarly journals Classification of Typical Tree Species in Laser Point Cloud Based on Deep Learning

2021 ◽  
Vol 13 (23) ◽  
pp. 4750
Author(s):  
Jianchang Chen ◽  
Yiming Chen ◽  
Zhengjun Liu
Keyword(s):  
Feature Extraction ◽  
Deep Learning ◽  
Tree Species ◽  
Classification Accuracy ◽  
Point Cloud ◽  
Laser Scanning ◽  
Species Classification ◽  
Tree Species Classification ◽  
Network Mapping

We propose the Point Cloud Tree Species Classification Network (PCTSCN) to overcome challenges in classifying tree species from laser data with deep learning methods. The network is mainly composed of two parts: a sampling component in the early stage and a feature extraction component in the later stage. We used geometric sampling to extract regions with local features from the tree contours since these tend to be species-specific. Then we used an improved Farthest Point Sampling method to extract the features from a global perspective. We input the intensity of the tree point cloud as a dimensional feature and spatial information into the neural network and mapped it to higher dimensions for feature extraction. We used the data obtained by Terrestrial Laser Scanning (TLS) and Unmanned Aerial Vehicle Laser Scanning (UAVLS) to conduct tree species classification experiments of white birch and larch. The experimental results showed that in both the TLS and UAVLS datasets, the input tree point cloud density and the highest feature dimensionality of the mapping had an impact on the classification accuracy of the tree species. When the single tree sample obtained by TLS consisted of 1024 points and the highest dimension of the network mapping was 512, the classification accuracy of the trained model reached 96%. For the individual tree samples obtained by UAVLS, which consisted of 2048 points and had the highest dimension of the network mapping of 1024, the classification accuracy of the trained model reached 92%. TLS data tree species classification accuracy of PCTSCN was improved by 2–9% compared with other models using the same point density, amount of data and highest feature dimension. The classification accuracy of tree species obtained by UAVLS was up to 8% higher. We propose PCTSCN to provide a new strategy for the intelligent classification of forest tree species.

Virtual laser scanning (VLS) in forestry – Investigating appropriate 3D forest representations for LiDAR simulations with HELIOS++

2021 ◽  
Author(s):  
Hannah Weiser ◽  
Lukas Winiwarter ◽  
Jannika Schäfer ◽  
Fabian Ewald Fassnacht ◽  
Katharina Anders ◽  
...  
Keyword(s):  
Point Cloud ◽  
Laser Scanning ◽  
Point Clouds ◽  
Projection Area ◽  
Species Classification ◽  
Voxel Size ◽  
Point Distribution ◽  
3D Point Clouds ◽  
Forest Models ◽  
Tree Metrics

&lt;p&gt;Virtual laser scanning (VLS) is a valuable method to complement expensive laser scanning data acquisition in the field. VLS refers to the simulation of LiDAR to create 3D point clouds from models of scenes, platforms and sensors mimicking real world acquisitions. In forestry, this can be used to generate training and testing data with complete ground truth for algorithms performing essential tasks such as tree detection or tree species classification. Furthermore, VLS allows for the in-depth investigation of the influence of different acquisition parameters on the point clouds and thus also the behaviour of algorithms, which is important when relating point cloud metrics to forest inventory variables. Finally, VLS can be used for acquisition planning and optimisation, as different configurations can be tested regarding their ability to create data of the required quality with minimal effort. For these purposes, we developed the open source Heidelberg LiDAR Operations Simulator HELIOS++ (written in C++) which is available on GitHub (https://github.com/3dgeo-heidelberg/helios), as a precompiled command line tool, and as Python package (pyhelios). HELIOS++ provides a high-fidelity framework for full 3D laser scanning simulations with multiple platforms and a flexible system to represent the scene. HELIOS++ models the beam divergence and supports the recording of the full waveform.&lt;/p&gt;&lt;p&gt;One important premise for the usefulness of VLS data is the use of an adequate 3D scene in the simulation. In this context, we conducted a study investigating point clouds simulated based on opaque voxel-based forest models computed from terrestrial laser scanning data using different voxel sizes. Coupling the LiDAR simulation with a database containing point clouds of single trees from terrestrial, UAV-borne and airborne acquisitions, allowed us to compare metrics derived from real and simulated data. Furthermore, by including the tree neighbourhood in the scene, we were able to consider occlusion effects between the trees.&lt;/p&gt;&lt;p&gt;We found that the voxel size is an important parameter, where values of e.g. 0.25 m lead to unrealistic occlusion effects of the mid- and understory, as only few gaps remain in the forest models through which the laser beam can pass. This results in fewer multiple returns, the vertical point distribution is shifted upwards, and tree metrics such as crown projection area and crown base height are estimated poorly. Smaller voxel sizes are therefore preferable, though the appropriate voxel size depends on the resolution of the input point cloud. With very small voxels, the voxel model may become too transparent. To achieve realistic simulations without the need for a high number of voxels we suggest variable downscaling of voxel cubes based on appropriate local metrics such as the plant area density. This approach decreases the computational requirements for the simulation, as fewer primitives are present in the scene. In our study, the use of such scaled voxels derived for a grid size of 0.25 m achieves equally and partly more reliable estimates of point cloud and tree metrics than regular voxels at fixed side lengths of 0.05 and 0.02 m.&lt;/p&gt;

Sign in / Sign up

Export Citation Format

Share Document