Cucumber seedlings segmentation network based on multi-view geometric graph encoder from 3D point clouds DOI Creative Commons
Yonglong Zhang,

Yaling Xie,

Jialuo Zhou

et al.

Plant Phenomics, Journal Year: 2024, Volume and Issue: 6

Published: Jan. 1, 2024

Plant phenotyping plays a pivotal role in observing and comprehending the growth development of plants. In phenotyping, plant organ segmentation based on 3D point clouds has garnered increasing attention recent years. However, using only geometric relationship features Euclidean space still cannot accurately segment measure To this end, we mine more propose network multiview graph encoder, called SN-MGGE. First, construct cloud acquisition platform to obtain cucumber seedling dataset, employ CloudCompare software annotate data. The GGE module is then designed generate features, including relationships shape structure, via encoder over hyperbolic spaces. Finally, semantic results are obtained downsampling operation multilayer perceptron. Extensive experiments dataset clearly show that our proposed SN-MGGE outperforms several mainstream networks (e.g., PointNet++, AGConv, PointMLP), achieving mIoU OA values 94.90% 97.43%, respectively. On basis results, 4 phenotypic parameters (i.e., height, leaf length, width, area) extracted through K-means clustering method; these very close ground truth,

Language: Английский

Evaluating Lorenz entropy for tropical forest discrimination using GEDI and supervised machine learning approach DOI
Nooshin Mashhadi, Arturo Sánchez‐Azofeifa

Ecological Indicators, Journal Year: 2025, Volume and Issue: 173, P. 113374 - 113374

Published: March 20, 2025

Language: Английский

Citations

0
Integrating DEM and Deep Learning for Forested Terrain Analysis: Enhancing Fire Risk Assessment Through Mountain Peak and Water System Extraction in Chongli District DOI Open Access

Yihui Wu,

Xueying Sun, Liang Qi

et al.

Forests, Journal Year: 2025, Volume and Issue: 16(4), P. 692 - 692

Published: April 16, 2025

Accurate fire risk assessment in forested terrain is crucial for effective disaster management and ecological conservation. This study innovatively proposes a novel framework that integrates Digital Elevation Models (DEMs) with deep learning techniques to enhance Chongli District. Our combines DEM data Faster Regions Convolutional Neural Networks (Faster R-CNN) CNN-based methods, breaking through the limitations of traditional approaches rely on manual feature extraction. It capable automatically identifying critical features, such as mountain peaks water systems, higher accuracy efficiency. DEMs provide high-resolution topographical information, which models leverage accurately identify delineate key geographical features. results show integration significantly improves by offering detailed precise analysis, thereby providing more reliable inputs behavior prediction. The extracted fundamental prediction, enable accurate predictions spread potential impact areas. not only highlights great combining geospatial advanced machine but also offers scalable efficient solution forest mountainous regions. Future work will focus expanding dataset include environmental variables validating model different areas further its robustness applicability.

Language: Английский

Citations

0
Research on eucalyptus individual tree segmentation and age estimation utilizing improved Mask R-CNN algorithm based on UAV stereo images DOI Creative Commons

Jirong Ding,

Li-Bi You,

Yehua Liang

et al.

Industrial Crops and Products, Journal Year: 2025, Volume and Issue: 230, P. 121073 - 121073

Published: April 23, 2025

Language: Английский

Citations

0
Automated 3D Segmentation of Plant Organs via the Plant-MAE: A Self-Supervised Learning Framework DOI Creative Commons
Kai Xie,

Chenxi Cui,

Xue Jiang

et al.

Plant Phenomics, Journal Year: 2025, Volume and Issue: unknown, P. 100049 - 100049

Published: May 1, 2025

Language: Английский

Citations

0
Forest Fire Prediction Based on Time Series Networks and Remote Sensing Images DOI Open Access
Yue Cao,

Xuanyu Zhou,

Yanqi Yu

et al.

Forests, Journal Year: 2024, Volume and Issue: 15(7), P. 1221 - 1221

Published: July 14, 2024

Protecting forest resources and preventing fires are vital for social development public well-being. However, current research studies on fire warning systems often focus extensive geographic areas like states, counties, provinces. This approach lacks the precision detail needed predicting in smaller regions. To address this gap, we propose a Transformer-based time series forecasting model aimed at improving accuracy of predictions areas. Our study focuses Quanzhou County, Guilin City, Guangxi Province, China. We utilized data from 2021 to 2022, along with remote sensing images ArcGIS technology, identify various factors influencing region. established dataset containing twelve factors, each labeled occurrences. By integrating these Transformer model, generated danger level prediction maps County. model’s performance is compared other deep learning methods using metrics such as RMSE, results reveal that proposed achieves higher (ACC = 0.903, MAPE 0.259, MAE 0.053, RMSE 0.389). demonstrates effectively takes advantage spatial background information periodicity significantly enhancing predictive accuracy.

Language: Английский

Citations

3
Concepts Towards Nation-Wide Individual Tree Data and Virtual Forests DOI Creative Commons

Matti Hyyppä,

Tuomas Turppa,

Heikki Hyyti

et al.

ISPRS International Journal of Geo-Information, Journal Year: 2024, Volume and Issue: 13(12), P. 424 - 424

Published: Nov. 26, 2024

Individual tree data could offer potential uses for both forestry and landscape visualization but has not yet been realized on a large scale. Relying 5 points/m2 Finnish national laser scanning, we present the design implementation of system producing, storing, distributing, querying, viewing individual data, in web browser game engine-mediated interactive 3D visualization, “virtual forest”. In our experiment, 3896 km2 airborne scanning point clouds were processed detection, resulting over 100 million trees detected, developed technical infrastructure allows containing 10+ billion (a rough number log-sized Finland) to be visualized same system. About 92% wider than 20 cm diameter at breast height (corresponding industrial log-size trees) detected using data. Obtained relative RMSE height, diameter, volume, biomass (stored above-ground carbon) levels 4.5%, 16.9%, 30.2%, 29.0%, respectively. The obtained bias are low enough operational add value current area-based inventories. By combining single-tree with open GIS datasets, virtual forest was produced automatically. A comparison against georeferenced panoramic images performed assess verisimilitude scenes, best results from sparse grown forests sites clear landmarks. Both online viewer can used improved decision-making multifunctional forestry. Based work, inventory is expected become Finland 2026 as part third program.

Language: Английский

Citations

2
Towards consistently measuring and monitoring habitat condition with airborne laser scanning and unmanned aerial vehicles DOI Creative Commons
W. Daniel Kissling, Yifang Shi, Wang Jin-hu

et al.

Ecological Indicators, Journal Year: 2024, Volume and Issue: 169, P. 112970 - 112970

Published: Dec. 1, 2024

Language: Английский

Citations

2
Forest aboveground biomass estimation based on spaceborne LiDAR combining machine learning model and geostatistical method DOI Creative Commons
Li Xu, Jinge Yu, Qingtai Shu

et al.

Frontiers in Plant Science, Journal Year: 2024, Volume and Issue: 15

Published: Dec. 11, 2024

Estimation of forest biomass at regional scale based on GEDI spaceborne LiDAR data is great significance for quality assessment and carbon cycle. To solve the problem discontinuous footprints, this study mapped different echo indexes in footprints to surface by inverse distance weighted interpolation method, verified influence number results. Random algorithm was chosen estimate spruce-fir combined with parameters provided 138 sample plots Shangri-La. The results show that: (1) By extracting numbers visualize it, revealed that a higher correlates denser distribution more pronounced stripe phenomenon. (2) prediction accuracy improves as decreases. group highest R 2 , lowest RMSE MAE footprint extracted every 100 shots, 10 shots had worst effect. (3) inverted random ranged from 51.33 t/hm 179.83 an average 101.98 . total value 3035.29 × 4 This shows will have certain impact mapping information presents methodological reference selecting appropriate derive various vertical structure ecosystems.

Language: Английский

Citations

2
YOLOTree-Individual Tree Spatial Positioning and Crown Volume Calculation Using UAV-RGB Imagery and LiDAR Data DOI Open Access

Taige Luo,

Shuyu Rao,

Wenjun Ma

et al.

Forests, Journal Year: 2024, Volume and Issue: 15(8), P. 1375 - 1375

Published: Aug. 6, 2024

Individual tree canopy extraction plays an important role in downstream studies such as plant phenotyping, panoptic segmentation and growth monitoring. Canopy volume calculation is essential part of these studies. However, existing methods based on LiDAR or UAV-RGB imagery cannot balance accuracy real-time performance. Thus, we propose a two-step individual volumetric modeling method: first, use RGB remote sensing images to obtain the crown information, then spatially aligned point cloud data height information automate volume. After introducing our method outperforms image-only 62.5% accuracy. The AbsoluteError decreased by 8.304. Compared with traditional 2.5D using only, proposed 93.306. Our also achieves fast vegetation over large area. Moreover, YOLOTree model more comprehensive than YOLO series detection, 0.81% improvement precision, ranks second whole for mAP50-95 metrics. We sample open-source TreeLD dataset contribute research migration.

Language: Английский

Citations

2
TreeLearn: A deep learning method for segmenting individual trees from ground-based LiDAR forest point clouds DOI Creative Commons
Jonathan Henrich, Jan van Delden, Dominik Seidel

et al.

Ecological Informatics, Journal Year: 2024, Volume and Issue: unknown, P. 102888 - 102888

Published: Nov. 1, 2024

Language: Английский

Citations

2