Non-destructive Estimates of Aboveground Biomass of Savanna Trees Using Quantitative Structure Models and Close-Range Photogrammetry.

Trees Forests and People, Journal Year: 2025, Volume and Issue: unknown, P. 100791 - 100791

Published: Feb. 1, 2025

Language: Английский

---

PointMM: Point Cloud Semantic Segmentation CNN under Multi-Spatial Feature Encoding and Multi-Head Attention Pooling

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(7), P. 1246 - 1246

Published: March 31, 2024

For the actual collected point cloud data, there are widespread challenges such as semantic inconsistency, density variations, and sparse spatial distribution. A network called PointMM is developed in this study to enhance accuracy of segmentation complex scenes. The main contribution involves two aspects: (1) Multi-spatial feature encoding. We leverage a novel encoding module learn multi-spatial features from neighborhood set obtained by k-nearest neighbors (KNN) space. This enhances network’s ability structures various samples more finely completely. (2) Multi-head attention pooling. multi-head pooling address limitations symmetric function-based pooling, maximum average terms losing detailed information. achieved aggregating attribute clouds, thereby enhancing transmit information comprehensively accurately. Experiments on publicly available datasets S3DIS ISPRS 3D Vaihingen demonstrate that effectively learns at different levels, while improving objects. Compared 12 state-of-the-art methods reported literature, outperforms runner-up 2.3% OA dataset, achieves third best performance both MioU dataset. Both achieve satisfactory balance between OA, F1, MioU.

Language: Английский

---

Quantifying Tree Structural Change in an African Savanna by Utilizing Multi-Temporal TLS Data

Remote Sensing, Journal Year: 2025, Volume and Issue: 17(5), P. 757 - 757

Published: Feb. 22, 2025

Structural changes in savanna trees vary spatially and temporally because of both biotic abiotic drivers, as well the complex interactions between them. Given this complexity, it is essential to monitor quantify woody structural savannas efficiently. We implemented a non-destructive approach based on Terrestrial Laser Scanning (TLS) Quantitative Structure Models (QSMs) that offers unique advantage investigating tree parameters, such volume branch length parameters have not been previously reported for trees. Leaf-off multi-scan TLS point clouds were acquired during dry season, using Riegl VZ1000 TLS, September 2015 October 2019 at Skukuza flux tower Kruger National Park, South Africa. These three-dimensional (3D) data covered an area 15.2 ha with average density 4270 points/m2 (0.015°) 1600 (0.025°) clouds, respectively. Individual segmentation was applied two comparative shortest-path algorithm LiDAR 360(v5.4) software. reconstructed optimized QSMs assessed Diameter Breast Height (DBH), height, crown area, volume, individual level. The DBH, trunk showed significant positive correlations (R2 > 0.80) scanning periods regardless difference number points matched opposite observed total branches, 1st-order length. As densities increased, computed also increased < 0.63) high relative difference. A 45% present identified damaged/felled (75 trees), lost estimated be 83.4 m3. results our study reconstruction algorithms TreeQSMs high-resolution datasets can used successfully structure are key understanding ecology given its dynamic nature accurately quantifying gains losses could arise from fire, drought, herbivory, other disturbances.

Language: Английский

---

Generic and Specific Models for Volume Estimation in Forest and Savanna Phytophysiognomies in Brazilian Cerrado

Plants, Journal Year: 2024, Volume and Issue: 13(19), P. 2769 - 2769

Published: Oct. 3, 2024

The Cerrado has high plant and vertebrate diversity is an important biome for conserving species provisioning ecosystem services. Volume equations in this are scarce because of their size physiognomic diversity. This study was conducted to develop specific volumetric models the phytophysiognomies Gallery Forest, Dry Forest Savannah, Savannah Woodland, a generic model forest formation. Twelve 10 m × (100 m²) (National Inventory) plots were used each phytophysiognomy at different sites (regions) Federal District (FD) where trees had diameter breast height (DBH; 1.30 m) ≥5 cm formations base (Db; 0.30 savanna formations. Their diameters heights measured, they cut cubed, volume tree obtained according Smalian methodology. Linear nonlinear adjusted. Criteria selection determined using correlation coefficients, standard error estimates, graphical analysis residues. They later validated by chi-square test. resultant indicated that fit ideal; however, formation exhibited similar performance could be extensive areas Cerrado, represent potential generalization. To further increase our understanding, research recommended development total areas.

Language: Английский

---