Article
Engineering, Electrical & Electronic
Jintao Li, Hangbin Wu, Xiaolong Cheng, Yuanhang Kong, Xufei Wang, Yanyi Li, Chun Liu
Summary: This article proposes a new method to extract street trees individually from mobile laser scanning point clouds. The method removes the ground and buildings through data preprocessing, further removes artificial poles that may overlap with street tree crowns through supervoxels region growing, and selects regions of interest (ROI) including street trees and understory vegetation. Then, the main branch part of each tree is separated from the ROI by nonphotosynthetic components clustering, and the remaining photosynthetic components in the ROI are segmented individually based on the individual clustering results.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2023)
Article
Engineering, Multidisciplinary
Jintao Li, Xiaojun Cheng, Zhihua Xiao
Summary: The study proposes a branch-trunk-constrained hierarchical clustering method to individually extract street trees from mobile laser scanning point clouds, achieving high extraction precision and recall rates.
Article
Environmental Sciences
Pengcheng Wang, Yong Tang, Zefan Liao, Yao Yan, Lei Dai, Shan Liu, Tengping Jiang
Summary: This study proposes a deep learning framework that combines semantic and instance segmentation to extract single road-side trees from vehicle-mounted mobile laser scanning (MLS) point clouds. The proposed method accurately extracts approximately 90% of the road-side trees and achieves better segmentation results than existing methods in two urban MLS point clouds. It provides a promising solution for ecological construction based on the Living Vegetation Volume (LVV) calculation of urban roads.
Article
Engineering, Electrical & Electronic
Jintao Li, Xiaojun Cheng, Zhenlun Wu, Wang Guo
Summary: The article introduces an over-segmentation-based uphill clustering method for individual extraction of urban street trees from mobile laser scanning data, which enhances extraction results in complex environments without relying heavily on tree trunks. Experimental results demonstrate the method's effectiveness and high accuracy in individual tree extraction.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2021)
Article
Remote Sensing
Tengping Jiang, Shan Liu, Qinyu Zhang, Xin Xu, Jian Sun, Yongjun Wang
Summary: This study presents an automated two-stage framework for accurate instance segmentation of street trees from point clouds. The proposed method outperforms existing methods in challenging cases and provides a practical solution for ecological assessment based on individual tree segmentation.
INTERNATIONAL JOURNAL OF APPLIED EARTH OBSERVATION AND GEOINFORMATION
(2023)
Article
Environmental Sciences
Xiaojuan Ning, Yishu Ma, Yuanyuan Hou, Zhiyong Lv, Haiyan Jin, Zengbo Wang, Yinghui Wang
Summary: In this paper, a method based on multi-feature enhancement and tree structure analysis is proposed to extract trees from scanned urban scenes. The method can effectively extract trees with multiple rows of occlusion and improve the accuracy of tree extraction.
Article
Agriculture, Multidisciplinary
Yang Wang, Xubing Yang, Li Zhang, Xijian Fan, Qiaolin Ye, Liyong Fu
Summary: In this study, a learning framework based on supervised data clustering is proposed to accurately and simultaneously obtain individual tree segmentation (ITS) and tree counting from two-dimensional (2D) top-view tree canopy crowns. A pixel-precision classifier is used to recognize tree pixels or superpixels, and a supervised clustering method is proposed to group the tree superpixels into individual trees. The results demonstrate that this method is superior to the state-of-the-art methods in both visualized ITS and numeric tree-counting results.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2023)
Review
Forestry
Mingxia Yang, Xiaolu Zhou, Zelin Liu, Peng Li, Jiayi Tang, Binggeng Xie, Changhui Peng
Summary: Understanding the biomass, characteristics, and carbon sequestration of urban forests is crucial for maintaining and improving the quality of life and ensuring sustainable urban planning. This review evaluates recent developments in urban forest research methods, compares the accuracy and efficiency of different methods, and identifies emerging themes in urban forest assessment.
Article
Chemistry, Analytical
Lino Comesana-Cebral, Joaquin Martinez-Sanchez, Henrique Lorenzo, Pedro Arias
Summary: Individual tree segmentation is important for forest management, and LiDAR technology has shown to be superior in this area. Using DBSCAN clustering and cylinder voxelization can improve the detection rate and accuracy of tree location identification.
Article
Plant Sciences
Wenjie Zhang, Baoguo Wu, Yi Ren, Guijun Yang
Summary: This study constructed an individual tree growth model for Chinese fir, taking into account the effects of competition and environment. The results showed that including competition and environmental factors improved the accuracy of the model.
Article
Engineering, Electrical & Electronic
Kai Xia, Hao Wang, Yinhui Yang, Xiaochen Du, Hailin Feng
Summary: This study developed a novel deep learning architecture for individual tree crown detection and parameter estimation, which was verified to be practical in complex urban scenes through testing. The architecture accurately identified tree crowns in various test scenarios, making it suitable for urban green space inventory management.
JOURNAL OF SENSORS
(2021)
Article
Geography, Physical
Kenneth Olofsson, Eva Lindberg, Johan Holmgren, Raul de Paula Pires, Henrik Jan Persson
Summary: This study proposes a car-mounted mobile laser scanner method for individual tree detection and stem attribute estimation, aiming to improve remote sensing-based forest inventories. The results show that this method can accurately retrieve tree information at different distances from the roadside, making it suitable for large-scale forest inventories.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2022)
Article
Environmental Sciences
Robin J. L. Hartley, Sadeepa Jayathunga, Peter D. Massam, Dilshan De Silva, Honey Jane Estarija, Sam J. Davidson, Adedamola Wuraola, Grant D. Pearse
Summary: Phenotyping has been used in horticultural industries for decades, but it was less accessible for tree breeders until recently when affordable and non-destructive technologies like mobile laser scanners became available. In this study, a high-density mobile laser scanner was used to derive phenotypic measurements from mature Pinus radiata, and the results showed strong agreement with field measurements. The findings suggest that MLS technology holds strong potential for advancing forest phenotyping and tree measurement, even in mature forests.
Article
Geography, Physical
Tengping Jiang, Yongjun Wang, Shan Liu, Qinyu Zhang, Lin Zhao, Jian Sun
Summary: As one of the most important components of urban space, road trees play a crucial role in assessing and improving urban environments. This study proposes an effective approach to segment individual trees from urban MLS data, using a two-step process of tree-point extraction and individual-tree segmentation. The proposed method achieves high accuracy in segmenting roadside trees and demonstrates good generalization ability.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2023)
Article
Environmental Sciences
Liqun Lin, Yangyan Deng, Man Peng, Longxiang Zhen, Shuwei Qin
Summary: This study investigated the multi-scale effects of indicators, including urban shadows, on urban thermal environment research (UTE) during mid-day hours. Results showed that temperature variations mid-day were better explained by smaller scales of landscape heterogeneity and larger scales of landscape consistency under shadow conditions, while temperature under direct sunlight was primarily influenced by larger scales. Trees significantly reduced temperature, while building area and height were significantly correlated with temperature. These findings provide crucial reference data for micro-scale UTE investigations during mid-day hours and offer new strategies for urban planning and design.
Article
Computer Science, Artificial Intelligence
Fa Zhu, Ye Ning, Xingchi Chen, Yongbin Zhao, Yining Gang
Summary: SVOR is an extension of support vector machine in ordinal regression problem, which requires more training time compared to SVC or SVR. By removing potential redundant constraints, the efficiency of SVOR can be significantly improved without degrading performance.
APPLIED SOFT COMPUTING
(2021)
Article
Geochemistry & Geophysics
Shaobo Xia, Sheng Nie, Pu Wang, Dong Chen, Sheng Xu, Cheng Wang
Summary: The proposed gap-based data dividing method aims to minimize the intersections between cutting lines and objects. Experimental results demonstrate that this method outperforms the baseline method in terms of visual inspection and cutting line quality.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Sheng Xu, Xuan Zhou, Weidu Ye, Qiaolin Ye
Summary: This study introduces a new augmented convolutional neural network (ACNN) for point cloud classification, which enhances local structure information. Adaptive learning of parameters and adjustment of smoothness play a significant role during the learning process, showcasing high robustness in point cloud processing.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Sheng Xu, Wen Han, Weidu Ye, Qiaolin Ye
Summary: In this letter, a semi-automatic method is proposed to extract desired segmentation contours by initializing, calculating internal and external forces, and solving deformation equations. Experimental results demonstrate the effectiveness of the method in terms of accuracy and consistency, outperforming selected primitive- and object-based methods.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Shaobo Xia, Sheng Xu, Ruisheng Wang, Jonathan Li, Guanghui Wang
Summary: This study presents a method to extract individual buildings from ALS point clouds using widely accessible polygonal footprints. The method can achieve high instance-level building mapping accuracy around 90% and future work will focus on improving classification errors in preprocessing, shape inconsistencies between point clouds and polygons, as well as building footprint delineation and updating in postprocessing.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Environmental Sciences
Sheng Xu, Xin Li, Jiayan Yun, Shanshan Xu
Summary: This paper proposes a four-step framework for tree skeleton extraction, achieving complete skeletons through optimizing paths and interpolating points, and providing an efficient solution for tree skeleton and structure study.
Article
Computer Science, Artificial Intelligence
Fa Zhu, Junbin Gao, Jian Yang, Ning Ye
Summary: Linear Discriminant Analysis (LDA) assumes samples from the same class are independently and identically distributed, which may lead to failure when there are multiple clusters within a class. This paper proposes a neighborhood linear discriminant analysis (nLDA) that defines scatter matrices based on a neighborhood of reverse nearest neighbors, eliminating the need for the i.i.d. assumption. Experimental results show that nLDA outperforms previous discriminators in terms of performance.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Artificial Intelligence
Wei Zheng, Shuo Chen, Zhenyong Fu, Fa Zhu, Hui Yan, Jian Yang
Summary: FSBUF is a novel embedded framework for feature selection that improves the generalization ability of traditional embedded methods by introducing an additional classifier for unselected features. Experimental results demonstrate its comprehensibility and superior performance.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Geochemistry & Geophysics
Xin Li, Xuan Zhou, Sheng Xu
Summary: In this study, we propose an innovative method for obtaining the complete skeletons and 3-D structures of individual trees using point clouds. Our method extracts individual trees from input street scenes and segments them into small successive pieces, with the centers of each piece serving as skeleton candidate points. Through interpolation based on Euclidean distance and orientation, the entire skeleton is obtained. Finally, a high-precision 3-D model of trees is constructed by cylindrically fitting the skeleton using optimized circular truncated cones. Experimental results show that our method is highly efficient and effective, achieving 98% accuracy and requiring less than 1 minute for reconstruction. Compared to other methods, our approach reduces the time by more than 95%.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Remote Sensing
Sheng Xu, Xin Li, Hongxin Yang, Shanshan Xu
Summary: This work aims to develop a deep learning framework for segmenting woods from tree point clouds. The authors propose a novel preprocessing layer called the projection layer, which transforms 3D point clouds into 2D points for the subsequent convolution process. The projection map is updated in the learning process to capture geometric structure information.
REMOTE SENSING LETTERS
(2023)
Article
Environmental Sciences
Zhouyang Hua, Sheng Xu, Yingan Liu
Summary: This paper proposes an individual tree segmentation method called Shadow-cut to extract the contours of street tree point clouds. The method includes using support vector machine for tree region separation, calculating the optimal projection, and using image segmentation algorithm to extract edges. Experiments demonstrate that this method achieves high accuracy and completeness on LiDAR data.
Article
Computer Science, Artificial Intelligence
Fa Zhu, Xingchi Chen, Shuo Chen, Wei Zheng, Weidu Ye
Summary: As a classical ordinal regression model, support vector ordinal regression (SVOR) finds parallel discriminant hyperplanes to maximize the minimal margins between different ranks. However, SVOR only considers minor patterns near the margin hyperplanes and ignores the contributions of other patterns. To address this issue, this paper proposes relative margin induced support vector ordinal regression (RMSVOR) models, which depict the margin between a pattern and a discriminant hyperplane based on relative margin information. Experimental results on various datasets show that RMSVOR outperforms previous ordinal regression models and canonical multi-class classification models.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Environmental Sciences
Tianyi Xie, Wen Han, Sheng Xu
Summary: This paper proposes YOLO-RS, an optimized object detection algorithm based on YOLOv4, which improves the detection accuracy and speed by introducing the ASFF structure, optimizing the SPP structure in YOLOv4, and introducing Lightnet.
Article
Geochemistry & Geophysics
Ling Xing, Hongyu Qu, Sheng Xu, Yao Tian
Summary: This article proposes a novel and effective method, named CLEGAN, for unpaired low-light image enhancement using self-similarity contrastive learning within a single GAN framework. The proposed method maximizes the mutual information between low-light and restored images.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Artificial Intelligence
Fa Zhu, Wenjie Zhang, Xingchi Chen, Xizhan Gao, Ning Ye
Summary: As a state-of-the-art supervised novelty detection model, SVM-SND can recognize novelty or the class of a test instance through a single model. However, maximizing minimum margin is not sufficient for ensuring classification generalization. This paper introduces margin distribution and proposes an LMD-SND model to enhance the performance of multi-class supervised novelty detection. Experimental results show that LMD-SND outperforms SVM-SND and achieves comparative performance with shallow and deep novelty detection models.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)