Winter Wheat Canopy Height Estimation Based on the Fusion of LiDAR and Multispectral Data

Agronomy, Journal Year: 2025, Volume and Issue: 15(5), P. 1094 - 1094

Published: April 29, 2025

Wheat canopy height is an important parameter for monitoring growth status. Accurately predicting the wheat can improve field management efficiency and optimize fertilization irrigation. Changes in characteristics of at different stages affect structure, leading to changes quality LiDAR point cloud (e.g., lower density, more noise points). Multispectral data capture these crop provide information about status wheat. Therefore, a method proposed that fuses features multispectral feature parameters estimate winter Low-altitude unmanned aerial systems (UASs) equipped with cameras were used collect from experimental fields during three key stages: green-up (GUS), jointing (JS), booting (BS). Analysis variance, variance inflation factor, Pearson correlation analysis employed extract significantly correlated height. Four estimation models constructed based on Optuna-optimized RF (OP-RF), Elastic Net regression, Extreme Gradient Boosting, Support Vector Regression models. The model training results showed OP-RF provided best performance across all coefficient determination values 0.921, 0.936, 0.842 GUS, JS, BS, respectively. root mean square error 0.009 m, 0.016 0.015 m. absolute 0.006 0.011 At same time, it was obtained fusing better than single type parameters. meet requirements prediction. These demonstrate fusion accuracy monitoring. provides valuable remote sensing phenotypic low densely planted crops also support assessment management.

Language: Английский

---

Development of a stream DTM generation method using vegetation and morphology composite filters with SfM point clouds

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: April 4, 2025

Developing method for generating accurate Digital Terrain Model (DTM) of streams is necessary due to the limitations traditional field survey methods, which are time-consuming and costly do not provide continuous data. The objective this study was develop an advanced high-quality DTM using Structure from Motion (SfM) A leveling conducted on four cross-sections Bokha stream in Icheon City, S. Korea, SfM-based produced Pix4Dmapper program Phantom 4 multispectral drone. Two vegetation filters (NDVI NDI) two morphological (ATIN CSF) were applied data, best filter combination identified based MAE RMSE analyses. integration NDVI CSF showed performance area, while a single application lowest bare area. effectiveness SfM eliminating waterfront confirmed, with overall 0.299 m 0.375 m. These findings suggest that DTMs riparian zones can be achieved efficiently limited budget time proposed methodology.

Language: Английский

---

Delving into the Potential of Deep Learning Algorithms for Point Cloud Segmentation at Organ Level in Plant Phenotyping

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(17), P. 3290 - 3290

Published: Sept. 4, 2024

Three-dimensional point clouds, as an advanced imaging technique, enable researchers to capture plant traits more precisely and comprehensively. The task of segmentation is crucial in phenotyping, yet current methods face limitations computational cost, accuracy, high-throughput capabilities. Consequently, many have adopted 3D cloud technology for organ-level segmentation, extending beyond manual 2D visual measurement methods. However, analyzing phenotypic using influenced by various factors such data acquisition environment, sensors, research subjects, model selection. Although the existing literature has summarized application this there been a lack in-depth comparison analysis at algorithm level. This paper evaluates performance deep learning models on clouds collected or generated under different scenarios. These include outdoor real planting scenarios indoor controlled environments, employing both active passive Nine classical were comprehensively evaluated: PointNet, PointNet++, PointMLP, DGCNN, PointCNN, PAConv, CurveNet, Point Transformer (PT), Stratified (ST). results indicate that ST achieved optimal across almost all environments albeit significant cost. transformer architecture points demonstrated considerable advantages over traditional feature extractors accommodating features longer ranges. Additionally, PAConv constructs weight matrices data-driven manner, enabling better adaptation scales organs. Finally, thorough discussion conducted from multiple perspectives, including construction, collection platforms.

Language: Английский

---

A One-Dimensional Light Detection and Ranging Array Scanner for Mapping Turfgrass Quality

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(12), P. 2215 - 2215

Published: June 19, 2024

The turfgrass industry supports golf courses, sports fields, and the landscaping lawn care industries worldwide. Identifying problem spots in is crucial for targeted remediation treatment. There have been attempts to create vehicle- or drone-based scanners predict quality; however, these methods often issues associated with high costs and/or a lack of accuracy due using colour rather than grass height (R2 = 0.30 0.90). new vehicle-mounted scanner system developed this study allows faster data collection more accurate representation quality compared currently available while being affordable reliable. Gryphon Turf Canopy Scanner (GTCS), low-cost one-dimensional LiDAR array, was used scan provide information about height, density, homogeneity. Tests were carried out over three months 2021, ground-truthing taken during same period. When utilizing non-linear regression, could percent bare field 0.47, root mean square error < 0.5 mm) an increase 8% random forest metric. potential environmental impact technology vast, as approach would reduce water, fertilizer, herbicide usage.

Language: Английский

---