TLS2trees: A scalable tree segmentation pipeline for TLS data

Methods in Ecology and Evolution, Journal Year: 2023, Volume and Issue: 14(12), P. 3083 - 3099

Published: Oct. 21, 2023

Abstract Above‐ground biomass (AGB) is an important metric used to quantify the mass of carbon stored in terrestrial ecosystems. For forests, this routinely estimated at plot scale (typically 1 ha) using inventory measurements and allometry. In recent years, laser scanning (TLS) has appeared as a disruptive technology that can generate more accurate assessment tree AGB; however, operationalising TLS methods had overcome number challenges. One such challenge segmentation individual trees from level point clouds are required estimate woody volume, often done manually (e.g. with interactive cloud editing software) be very time consuming. Here we present TLS2trees , automated processing pipeline set Python command line tools aims redress bottleneck. consists existing new specifically designed horizontally scalable. The demonstrated on 7.5 ha data captured across 10 plots seven forest types; open savanna dense tropical rainforest. A total 10,557 segmented : these compared 1281 trees. Results indicate performs well, particularly for larger (i.e. cohort largest comprise 50% volume), where plot‐wise volume bias ±0.4 m 3 %RMSE 60%. Segmentation performance decreases smaller trees, example DBH ≤10 cm; reasons suggested including semantic step. increasing. It fully utilise activities monitoring, reporting verification or reference satellite missions pipeline, required. To facilitate improvements well modification other modes mobile UAV scanning), free open‐source software.

Language: Английский

---

UAV-derived photogrammetric point clouds and multispectral indices for fuel estimation in Mediterranean forests

Remote Sensing Applications Society and Environment, Journal Year: 2023, Volume and Issue: 31, P. 100997 - 100997

Published: May 25, 2023

Sensors attached to unmanned aerial vehicles (UAVs) allow estimating a large number of forest attributes related fuels. This study assesses photogrammetric point clouds and multispectral indices obtained from fixed-wing UAV for the classification Prometheus fuel types in 82 plots Aragón (NE Spain). Images captured by an RGB camera sensor allowed generating high density (RGB: 3000 points/m2; multispectral: 85 points/m2), which were normalized using alternatively Digital Elevation Model (DEM) 0.5, 1, 2 m resolution. A set structural textural variables derived cloud heights, latter, gray-level co-occurrence matrix (GLCM) approach was used. Multispectral images also used create seven spectral vegetation indices. The most relevant structural, textural, introduce into models selected Dunn's test, included: height at 50th percentile, coefficient variation percentage returns above 4 m, mean dissimilarity, Green Chlorophyll Index. Three different data samples introduced models: i) (RGB sample); ii) (MS iii) plus variable (integrated sample). After comparing three machine learning techniques (Random Forest, Linear Radial Support Vector Machine), best results with Random Forest k-fold cross-validation (k-10) integrated sample 0.5 DEM resolution (overall accuracy = 71%). successfully identified main fire carriers (i.e., shrubs or trees) confusions mainly located within same dominant stratum, especially 3 6. These demonstrate ability imagery classify fuels Mediterranean environments when are combined.

Language: Английский

---

Tree Crown Segmentation and Diameter at Breast Height Prediction Based on BlendMask in Unmanned Aerial Vehicle Imagery

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(2), P. 368 - 368

Published: Jan. 16, 2024

The surveying of forestry resources has recently shifted toward precision and real-time monitoring. This study utilized the BlendMask algorithm for accurately outlining tree crowns introduced a Bayesian neural network to create model linking individual crown size with diameter at breast height (DBH). outlines shapes contours, outperforming traditional watershed algorithms in segmentation accuracy while preserving edge details across different scales. Subsequently, constructs predicting DBH from measured area, providing essential data managing forest conducting biodiversity research. Evaluation metrics like rate, recall F1-score, mAP index comprehensively assess method’s performance regarding density. demonstrated higher 0.893 compared algorithm’s 0.721 based on experimental results. Importantly, effectively handles over-segmentation problems Moreover, adjusting parameters during execution allows flexibility achieving diverse image effects. addresses challenges builds area using network. average discrepancies between calculated Ginkgo biloba, Pinus tabuliformis, Populus nigra varitalica were 0.15 cm, 0.29 0.49cm, respectively, all within acceptable error margin 1 cm. BlendMask, besides its effectiveness segmentation, proves useful various vegetation classification tasks broad-leaved forests, coniferous grasslands. With abundant training ongoing parameter adjustments, attains improved accuracy. new approach shows great potential real-world use, offering crucial resources, research, related fields, aiding decision-making processes.

Language: Английский

---

Forest Canopy Fuel Loads Mapping Using Unmanned Aerial Vehicle High-Resolution Red, Green, Blue and Multispectral Imagery

Forests, Journal Year: 2024, Volume and Issue: 15(2), P. 225 - 225

Published: Jan. 24, 2024

Canopy fuels determine the characteristics of entire complex forest due to their constant changes triggered by environment; therefore, development appropriate strategies for fire management and risk reduction requires an accurate description canopy fuels. This paper presents a method mapping spatial distribution fuel loads (CFLs) in alignment with natural variability three-dimensional distribution. The approach leverages object-based machine learning framework UAV multispectral data photogrammetric point clouds. proposed was developed mixed protected area “Sierra de Quila”, Jalisco, Mexico. Structural variables derived from clouds, along spectral information, were used Random Forest model accurately estimate CFLs, yielding R2 = 0.75, RMSE 1.78 Mg, average Biasrel 18.62%. volume most significant explanatory variable, achieving mean decrease impurity values greater than 80%, while combination texture vegetation indices presented importance close 20%. Our modelling enables estimation accounting ecological context that governs dynamics variability. high precision achieved, at relatively low cost, encourages updating maps enable researchers managers streamline decision making on management.

Language: Английский

---

Peering through the thicket: Effects of UAV LiDAR scanner settings and flight planning on canopy volume discovery

International Journal of Applied Earth Observation and Geoinformation, Journal Year: 2022, Volume and Issue: 114, P. 103056 - 103056

Published: Nov. 1, 2022

Unoccupied aerial vehicle laser scanning (UAV-LS) has been increasingly used for forest structure assessment in recent years due to the potential directly estimate individual tree attributes and availability of commercial solutions. However, standardised procedures campaign planning are still largely missing. This study investigated scanner properties flight provide recommendations on minimising canopy occlusion thereby maximise exploration volume. A involving two UAV-LS systems was conducted over a dense, wet tropical at Paracou research station (French Guiana). Four experiments were conducted, analysed derived. First, pulse repetition rate (PRR) should be least 100 kHz per 1 m s−1 speed based 360° FOV middle strata (5 20 m). Higher PRR beneficial lower (<5 m) but would need increased exponentially achieve linear improvement. Alternatively, could reduced within constraints given by inertial measurement unit (IMU), increase time. Second, maximum range identified as proxy power, which positively impacts exploration. particularly case when using multi-return capabilities. No saturation observed increasing suggesting that this is currently limiting factor. Additionally, smaller beam divergence width plausible reasons better upper just below top canopy. Third, off-nadir angles up 20° found result similar occlusions, practical 40° dense forest. number might larger open canopies. with viewing geometries focus pulses downwards optimal ranges preferred. Fourth, different horizontal directions mission favours minimisation occlusion. minimum suggested. specific yaw not possible predict before flight. Therefore, including multiple ensures coverage all configurations. Many these features can optimised independently from each other, considered acquisition new planning. These results support establishment general guidelines investment assessment.

Language: Английский

---

Automatic tree detection and attribute characterization using portable terrestrial lidar

Trees, Journal Year: 2023, Volume and Issue: 37(3), P. 963 - 979

Published: March 31, 2023

Abstract Key message This study details a methodology to automatically detect the positions of and dasometric information about individual Eucalyptus trees from point cloud acquired with portable LiDAR system. Currently, implementation laser scanners (PLS) in forest inventories is being studied, since they allow for significantly reduced field-work time costs when compared traditional inventory methods other systems. However, it has been shown that their operability efficiency are dependent upon species assessed, therefore, there need more research assessing different types stands species. Additionally, few studies have conducted stands, one tree genus most commonly planted around world. In this study, PLS system was tested globulus stand obtain metrics trees. An automatic data (individual positions, DBH, diameter at heights, height trees) developed using public domain software. The results were obtained static terrestrial scanner (TLS). able identify 100% present both TLS clouds. For cloud, RMSE DBH 0.0716, 0.176. 3.415 m, while 10.712 m. demonstrates applicability systems estimation adult stands.

Language: Английский

---

Individual Tree-Scale Aboveground Biomass Estimation of Woody Vegetation in a Semi-Arid Savanna Using 3D Data

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(2), P. 399 - 399

Published: Jan. 19, 2024

Allometric equations are the most common way of assessing Aboveground biomass (AGB) but few exist for savanna ecosystems. The need accurate estimation AGB has triggered an increase in amount research towards 3D quantification tree architecture through Terrestrial Laser Scanning (TLS). Quantitative Structure Models (QSMs) trees have been described as way. However, accuracy using QSMs yet to be established savanna. We implemented a non-destructive method based on TLS and QSMs. Leaf-off multi scan point clouds were acquired 2015 Kruger National Park, South Africa Riegl VZ1000. data covered 80.8 ha with average density 315.3 points/m2. Individual segmentation was applied comparative shortest-path algorithm, resulting 1000 trees. As 31 failed reconstructed, we reconstructed optimized 969 computed volume converted wood 0.9. TLS-derived compared from three allometric equations. best modelling results had RMSE 348.75 kg (mean = 416.4 kg) Concordance Correlation Coefficient (CCC) 0.91. Optimized model repetition gave robust estimates given by low coefficient variation (CoV 19.9% 27.5%). limitations can addressed application high-density data. Our study shows that vegetation modelled clouds. this key understanding ecology, its complex dynamic nature.

Language: Английский

---

Ground-based calibration for remote sensing of biomass in the tallest forests

Forest Ecology and Management, Journal Year: 2024, Volume and Issue: 561, P. 121879 - 121879

Published: April 13, 2024

Forest biomass is a critical component of the terrestrial carbon cycle. The highest-biomass forests are those dominated by tallest species, Sequoia sempervirens. We use ground-based measurements and allometric equations to estimate tree in primary (40–42° N latitude) recently subjected spaceborne airborne laser scanning (GEDI ALS, respectively), we develop new allometry using GEDI ALS predictors. best equation for (live + dead) aboveground these forests, which based on 88th percentile relative height pulse return energy (N = 200 pulses, R2 0.37, RMSE 48%), predicts average per-hectare values statistically indistinguishable from predicted previously published (916 ± 74 vs. 928 11 Mg ha−1, mean 1 SE). equation, crown size approximate objects (dominant trees plus subordinates) segmented lidar datasets 503 segments, 0.64, 49%), significantly higher live than across 37465 ha forest surveyed (1384 77 885 73 Underestimation occurs because alone poor predictor forests. also moderately underestimates biomass, part neither nor can adequately account giant trunks. Despite shortcomings, demonstrate how hierarchy be used map distribution with global maximum density. Among seven reserves, estimated exceeds 2000 ha−1 three, ultrahigh-biomass (> 3000 ha−1) hectares sparsely distributed (1%) largest concentration occurring low-elevation alluvial terraces (460 ha) Humboldt Redwoods State Park. ALS-predicted provides realistic context-specific benchmarks ongoing restoration management logged inside reserves.

Language: Английский

---

Biomass estimation of abandoned orange trees using UAV-SFM 3D points

International Journal of Applied Earth Observation and Geoinformation, Journal Year: 2024, Volume and Issue: 130, P. 103931 - 103931

Published: May 23, 2024

In smallholder areas, the abandonment of orchards is a recent phenomenon with socioeconomic and environmental consequences. Biomass estimation monitoring these areas essential to analyze their influence on CO2 balance quantify carbon pools. current context energy supply uncertainties considering demanding use alternative sources, quantification fruit tree biomass in abandoned question great interest. this study, above orange trees was estimated using parameters calculated from 3D points derived images captured by UAV applying Structure Motion (SfM) technique. From data, canopy height model used apply developed crown contour detection algorithm. Using information points, area, diameter, length, maximum height, minimum mean standard deviation point heights were for set 36 felled weighted trees. Stepwise regression estimate values. All previously reported variables included. The area parameter produced most accurate R2, RMSE % values ​​of 0.85, 10.165 kg 19.56 %, respectively. These results demonstrate potential UAV-SfM-derived clouds above-ground trees, relevant analysis biofuel production.

Language: Английский

---

Treegraph: tree architecture from terrestrial laser scanning point clouds

Remote Sensing in Ecology and Conservation, Journal Year: 2024, Volume and Issue: unknown

Published: June 3, 2024

Abstract Accurate quantification of tree architecture is critical to interpreting the growth, health and functioning trees forests. Terrestrial laser scanning (TLS) offers millimetre‐level point cloud data, but current approaches 3D reconstruction from TLS clouds primarily focus on retrieving total volume at scale for aboveground biomass (AGB) estimation. Few methods have been designed specifically provide architectural properties, including branch‐level morphology topology, rather than AGB; derived topological traits tended be a compromise, secondary importance volume. We present Treegraph , new approach explicitly retrieve multiple scales, whole down individual branches internodes, using data with limited assumptions about form. It provides morphological such as branch length diameter, alongside parent–daughter connections furcation (branching) number order. compare ‐derived manual measurements eight destructively harvested trees, yielding RMSE values 0.60 m (5.96%) length, 2.99 cm (33.45%) 0.46 (19.38%) 0.08 (33.16%) internode respectively. In broader application 603 tropical, temperate urban forests, we demonstrate that support testing structure‐related metabolic scaling theories. Testing over 10 in diameter across 18 657 branching nodes shows exponents deviate WBE predictions, exhibiting area‐preserving behaviour while displaying asymmetry daughter branches. Available open‐source Python software, fine‐level network information, promoting improved insights into structure function. This data‐driven reduces need empirical heuristic parameters, which has potential advancing large‐scale ecological studies architecture.

Language: Английский

---