Tree Classification in Complex Forest Point Clouds Based on Deep Learning

Zou, Xinhuai; Cheng, Ming; Wang, Cheng; Xia, Yan; Li, Jonathan

doi:10.1109/lgrs.2017.2764938

Cited by 97 publications

(65 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Specifically, Hu and Yuan [44] suggest that DL-based algorithms can outperform the current methods that are most commonly used for ground return classification. Others have investigated the classification of features in 3D space represented as point clouds [45][46][47].…”

Section: Deep Learningmentioning

confidence: 99%

Mapping the Topographic Features of Mining-Related Valley Fills Using Mask R-CNN Deep Learning and Digital Elevation Data

2020

View full text Add to dashboard Cite

Modern elevation-determining remote sensing technologies such as light-detection and ranging (LiDAR) produce a wealth of topographic information that is increasingly being used in a wide range of disciplines, including archaeology and geomorphology. However, automated methods for mapping topographic features have remained a significant challenge. Deep learning (DL) mask regional-convolutional neural networks (Mask R-CNN), which provides context-based instance mapping, offers the potential to overcome many of the difficulties of previous approaches to topographic mapping. We therefore explore the application of Mask R-CNN to extract valley fill faces (VFFs), which are a product of mountaintop removal (MTR) coal mining in the Appalachian region of the eastern United States. LiDAR-derived slopeshades are provided as the only predictor variable in the model. Model generalization is evaluated by mapping multiple study sites outside the training data region. A range of assessment methods, including precision, recall, and F1 score, all based on VFF counts, as well as area- and a fuzzy area-based user’s and producer’s accuracy, indicate that the model was successful in mapping VFFs in new geographic regions, using elevation data derived from different LiDAR sensors. Precision, recall, and F1-score values were above 0.85 using VFF counts while user’s and producer’s accuracy were above 0.75 and 0.85 when using the area- and fuzzy area-based methods, respectively, when averaged across all study areas characterized with LiDAR data. Due to the limited availability of LiDAR data until relatively recently, we also assessed how well the model generalizes to terrain data created using photogrammetric methods that characterize past terrain conditions. Unfortunately, the model was not sufficiently general to allow successful mapping of VFFs using photogrammetrically-derived slopeshades, as all assessment metrics were lower than 0.60; however, this may partially be attributed to the quality of the photogrammetric data. The overall results suggest that the combination of Mask R-CNN and LiDAR has great potential for mapping anthropogenic and natural landscape features. To realize this vision, however, research on the mapping of other topographic features is needed, as well as the development of large topographic training datasets including a variety of features for calibrating and testing new methods.

show abstract

Section: Deep Learningmentioning

confidence: 99%

Mapping the Topographic Features of Mining-Related Valley Fills Using Mask R-CNN Deep Learning and Digital Elevation Data

2020

View full text Add to dashboard Cite

show abstract

“…Zhong et al [30] significantly improved the separation of overlapping tree crowns in the upper canopy layer, but could not distinguish individual trees in vertically layered forest structures. Zou et al [31] did not extract the complete trees, but only point cloud segments within a defined radius around the stem position that were intended to be used for species classification. Some studies used mobile laser scanning data (MLS) from a moving vehicle to detect and isolate single trees [32][33][34].…”

Section: Introductionmentioning

confidence: 99%

Constrained Spectral Clustering of Individual Trees in Dense Forest Using Terrestrial Laser Scanning Data

Heinzel

Huber

2018

Remote Sensing

View full text Add to dashboard Cite

Abstract:The present study introduces an advanced method for 3D segmentation of terrestrial laser scanning data into single tree clusters. It intentionally tackled difficult forest situations with dense and structured tree formations, which inventory practitioners are often faced with. The strongly interlocking tree crowns of different sizes and in different layers characterized the test conditions of close to nature forest plots. Volumetric 3D images of the plots were derived from the original point cloud data. A clustering method with automatically derived priors focused on the segmentation of these images by global optimization. Therefore, each image was segmented as a whole and partitioned into individual tree objects using a combination of state-of-the-art techniques. Multiple steps were combined in a workflow that included a morphological detection of the tree stems, the construction of a similarity graph from the image data, the computation of the eigenspectrum which was weighted with the tree stem priors and the final labelling of the transformed data points in a Markov Random Field framework. The detected trees were verified by number and position which allowed for comparison with other studies. Additionally, for a subset of the data, we provided a detailed verification of the three-dimensional extent of the complete trees. The detection rate by number and position was 97.40% for major trees with a stem diameter at breast height (DBH) ≥ 12 cm and 84.62% for regeneration trees with a DBH < 12 cm. The three-dimensional extent of the detected trees resulted in an average producer's accuracy of 93.66% and a user's accuracy of 94.06%. Overall, these numbers confirm the capacity of the method for accurate segmentation of strongly layered and understory trees. Future studies could test the method on wider areas with large scale data and different forest types in order to determine its general transferability.

show abstract

“…A promising classification performance was achieved on the test LiDAR data. Thus, to further demonstrate the superior performance of the 3D-CNN model, we compared it with the following two methods: DBM-based method (Guan et al 2014) and DBN-based method (Zou et al 2017). For the DBM-based method, first, a tree point cloud is converted into a waveform representation, which well reflects the geometrical properties of a tree.…”

Section: Comparative Experimentsmentioning

confidence: 99%

“…Guan et al (2014) proposed a Deep Boltzmann Machines (DBMs) based tree classification method, which classify ten tree species from the tree waveform representation, reflecting tree geometric structures in mobile LiDAR data. Zou et al (2017) proposed a deep belief network (DBN) model based tree classification method, in which the generated high-level features were used in a softmax classifier in the tree species classification step. However, these methods were applied deep learning algorithms to generate low-level or high-level features, which were input into a machine learning classifier later.…”

Section: Introductionmentioning

confidence: 99%

3d-CNN Based Tree Species Classification Using Mobile Lidar Data

Guan¹,

Yu²,

Yan³

et al. 2019

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

Our work addresses the problem of classifying tree species from mobile LiDAR data. The work is a two step-wise strategy, including tree segmentation and tree species classification. In the tree segmentation step, a voxel-based upward growing filtering is proposed to remove terrain points from the mobile laser scanning data. Then, individual trees are segmented via a Euclidean distance clustering approach and Voxel-based Normalized Cut (VNCut) segmentation approach. In the tree species classification, a voxel-based 3D convolutional neural network (3D-CNN) model is developed based on intensity information. A road section data acquired by a RIEGL VMX-450 system are selected for evaluating the proposed tree classification method. Qualitative analysis shows that our algorithm achieves a good performance.

show abstract

Tree Classification in Complex Forest Point Clouds Based on Deep Learning

Cited by 97 publications

References 16 publications

Mapping the Topographic Features of Mining-Related Valley Fills Using Mask R-CNN Deep Learning and Digital Elevation Data

Mapping the Topographic Features of Mining-Related Valley Fills Using Mask R-CNN Deep Learning and Digital Elevation Data

Constrained Spectral Clustering of Individual Trees in Dense Forest Using Terrestrial Laser Scanning Data

3d-CNN Based Tree Species Classification Using Mobile Lidar Data

Contact Info

Product

Resources

About