Evaluation of Machine Learning Regression Techniques for Estimating Winter Wheat Biomass Using Biophysical, Biochemical, and UAV Multispectral Data

Drones, Год журнала: 2024, Номер 8(7), С. 287 - 287

Опубликована: Июнь 26, 2024

Crop above-ground biomass (AGB) estimation is a critical practice in precision agriculture (PA) and vital for monitoring crop health predicting yields. Accurate AGB allows farmers to take timely actions maximize yields within given growth season. The objective of this study use unmanned aerial vehicle (UAV) multispectral imagery, along with derived vegetation indices (VI), plant height, leaf area index (LAI), nutrient content ratios, predict the dry (g/m2) winter wheat field southwestern Ontario, Canada. This assessed effectiveness Random Forest (RF) Support Vector Regression (SVR) models ABG from 42 variables. RF consistently outperformed SVR models, top-performing model utilizing 20 selected variables based on their contribution increasing node purity decision trees. achieved an R2 0.81 root mean square error (RMSE) 149.95 g/m2. Notably, included combination MicaSense bands, VIs, levels, height. significantly all other that relied solely UAV data or content. insights gained can enhance management AGB, leading more effective yield predictions management.

Язык: Английский

---

Ground-Based Hyperspectral Estimation of Maize Leaf Chlorophyll Content Considering Phenological Characteristics

Remote Sensing, Год журнала: 2024, Номер 16(12), С. 2133 - 2133

Опубликована: Июнь 13, 2024

Accurately measuring leaf chlorophyll content (LCC) is crucial for monitoring maize growth. This study aims to rapidly and non-destructively estimate the LCC during four critical growth stages investigate ability of phenological parameters (PPs) LCC. First, spectra were obtained by spectral denoising followed transformation. Next, sensitive bands (Rλ), indices (SIs), PPs extracted from all at each stage. Then, univariate models constructed determine their potential independent estimation. The multivariate regression (LCC-MR) built based on SIs, SIs + Rλ, Rλ after feature variable selection. results indicate that our machine-learning-based LCC-MR demonstrated high overall accuracy. Notably, 83.33% 58.33% these showed improved accuracy when successively introduced SIs. Additionally, model accuracies milk-ripe tasseling outperformed those flare–opening jointing under identical conditions. optimal was created using XGBoost, incorporating SI, PP variables R3 These findings will provide guidance support management.

Язык: Английский

---

Surface displacement measurement and modeling of the Shah-Gheyb salt dome in southern Iran using InSAR and machine learning techniques

International Journal of Applied Earth Observation and Geoinformation, Год журнала: 2024, Номер 132, С. 104016 - 104016

Опубликована: Июль 11, 2024

Salt domes play a crucial role in hydrocarbon storage, underground construction, solution mining, and mineralization. Therefore, deformation monitoring is essential for analyzing the kinematics impact of salt domes. This study aims to measure temporal displacements Shah-Gheyb dome from 2016 2019 2020 2022 using New Small Baseline Subset (NSBAS) Interferometric Synthetic Aperture Radar (InSAR) technique predict future through machine learning models. A total 14 data layers, including topography, remote sensing, hydrology, geology group were used Machine Learning (ML). Random Forest Regression (RFR) Support Vector (SVR) models employed project both East-West (E-W) Up-Down (U-D) components 29 scenarios. In E-W direction, exhibits displacement rate 39 mm/year, while U-D it varies between −18 +6 mm/year. ML predictions SAR interferometry processing results period 2020–2022 validated Root Mean Square Error (RMSE) correlation coefficient (R). The RFR model demonstrated lowest RMSE 1.9 mm component, achieving maximum R-value 97.3 %. For was 2.8 mm, with an 55.8 Evaluation predictive performance comparison InSAR outcomes indicated that predicted along directions greater accuracy than SVR. Furthermore, comparing by two perpendicular profiles confirmed model's precision.

Язык: Английский

---

Synergistic approach for land use and land cover dynamics prediction in Uttarakhand using cellular automata and Artificial neural network

GEOMATICA, Год журнала: 2024, Номер 76(2), С. 100017 - 100017

Опубликована: Авг. 10, 2024

Alterations in Land use and cover (LULC) stand out as a key catalyst for shifts global climate patterns, environmental conditions, ecological dynamics. In order to further enhance our comprehension of the effects variability on environment, Remote sensing GIS analytical approaches have been thoroughly explored are reflected an imperative vision. Thus, objective this study is model Uttarakhand's LULC pattern 2032 analyse changes trend between 1992 2022. change mapping was conducted utilizing semi-automated hybrid classification approach high level accuracy which integrates both Maximum likelihood Object based image analysis techniques Landsat datasets. The machine learning Cellular automata Artificial neural networks (CA-ANN) within MOLUSCE plugin QGIS applied future patterns. assessment results showed that overall years 1992, 2002, 2012, 2022 96.94 %, 97.77 98.61 % 98.87 respectively, kappa statistics coefficient 0.92, 0.95, 0.94 0.95 respectively. simulated projected map implies substantially accuracy, with Kappa value 0.77 85.39 correctness. Then, year predicted using CA-ANN. observed alterations significant, characterized by augmentation built-up areas, open land, water bodies, alongside decline snow-covered regions, vegetation cover. Whereas, slight increase seen Forested areas. Planners policy makers aiming accomplish more sustainable efficient management environment will find over prolonged period time be useful asset optimal land planning.

Язык: Английский

---

Crop Yield Prediction Using Machine Learning: An Extensive and Systematic Literature Review

Smart Agricultural Technology, Год журнала: 2024, Номер unknown, С. 100718 - 100718

Опубликована: Дек. 1, 2024

Язык: Английский

---

Inversion of winter wheat leaf area index from UAV multispectral images: classical vs. deep learning approaches

Frontiers in Plant Science, Год журнала: 2024, Номер 15

Опубликована: Март 14, 2024

Precise and timely leaf area index (LAI) estimation for winter wheat is crucial precision agriculture. The emergence of high-resolution unmanned aerial vehicle (UAV) data machine learning techniques offers a revolutionary approach fine-scale LAI at the low cost. While has proven valuable estimation, there are still model limitations variations that impede accurate efficient inversion. This study explores potential classical models deep estimating using multispectral images acquired by drones. Initially, texture features vegetation indices served as inputs partial least squares regression (PLSR) random forest (RF) model. Then, ground-measured were combined to invert LAI. In contrast, this also employed convolutional neural network (CNN) solely utilizes cropped original image estimation. results show outperform in terms correlation analysis with accuracy. However, highest accuracy achieved combining both conventional methods. Among three models, CNN yielded ( R 2 = 0.83), followed RF 0.82), PLSR exhibited lowest 0.78). spatial distribution values estimated similar, whereas differs significantly from first two models. achieves rapid findings can serve reference real-time growth monitoring field management practices.

Язык: Английский

---

An improved deep learning approach for detection of maize tassels using UAV-based RGB images

International Journal of Applied Earth Observation and Geoinformation, Год журнала: 2024, Номер 130, С. 103922 - 103922

Опубликована: Май 22, 2024

The emergence of maize tassels is the turning vegetative stage to reproductive (Zea mays L.), which critical for estimating grain yields. Recent advances in unmanned aerial vehicles (UAVs) remote sensing and deep learning-based object detection technique have provided a new approach detecting tassels. Meanwhile, there still exists challenges accurate due uncertainties complex field environment. existing networks fall accurately overlapping or small-scale tassels, as well exhibiting insufficient capability strong lighting conditions. Furthermore, current dataset exhibits limited temporal scope, unable encompass whole tasseling progress. In this study, we proposed FMTS dataset, designed novel called RESAM-YOLOv8n (Residual Spatial Attention Module-You Only Look Once v8n), introducing RESAM module training network with larger input image sizes. These enabled focus on important tassel features neglect irrelevant information, thereby enhancing its capability. was trained evaluated using mAP0.5, mAP0.75, Recall, Precision, F1 were 95.74 %, 66.70 89.28 95.59 92.00 respectively. counting number R2 value between network's ground truth reached 0.976, low RMSE 1.56 results showed better performance network, providing an effective method identifying

Язык: Английский

---

Surface water area dynamics of the major lakes of Ethiopia (1985–2023): A spatio-temporal analysis

International Journal of Applied Earth Observation and Geoinformation, Год журнала: 2024, Номер 132, С. 104007 - 104007

Опубликована: Июль 13, 2024

Ethiopian lakes face multiple ecosystem threats from human and natural factors, including climate change. The long-term dynamics of water surface area in many remain unknown. This study presents the first comprehensive account major lakes' dynamics, providing essential information for prioritizing hydrological conservation strategies. We extracted time-series Landsat images Google Earth Engine (GEE), using Modified Normalized Water Index (MNDWI) machine learning (Random Forest) methods. Random Forest (RF) outperformed MNDWI across lakes, especially when body spectral characteristics became complex. A comparison results with Joint Research Center (JRC) Global Surface Mapping Layers showed a good agreement. Long-term both common distinct trends. Common trends included: (a) similar recession expansion periodicities despite being different climatic zones, (b) an increase inter-annual variability after 2000, (c) abrupt point time (2000/2001). Distinct (i) decline five (Abiyata, Langano, Shalla, Chamo, Hayq), (ii) three (Ziway, Hawassa, Abaya), (iii) non-homogeneous two (Tana Hardibo). These show interplay phenomena lake-specific factors. Climate factors govern lake change (decreasing/increasing) cycle periods. magnitudes changes recovery rates each vary based on recommend efforts high storage degradation risks, as indicated by significant loss over historical period persistent declining includes Abiyata, Hayq, Hardibo, order priority.

Язык: Английский

---

Machine learning-based modeling of chl-a concentration in Northern marine regions using oceanic and atmospheric data

Frontiers in Marine Science, Год журнала: 2024, Номер 11

Опубликована: Июль 25, 2024

Chl-a concentration is one of the key characteristics marine areas related to photosynthesis, along with oxygen levels and water salinity. Most studies focus on estimating chl-a in closed bodies, rivers, coastal tropical temperate Earth belts are therefore limited specific regions also require direct measurements chemical analysis obtain precise information about environmental conditions. Remote sensing techniques spatial modeling aim offer tools for rapid global climate ecological changes. In this study, we develop a machine learning (ML)-based approach estimate chlorophyll-a when satellite data unavailable. To provide physical parameters that may influence predicted variable (chl-a concentration), combined observations from MODIS geophysical Weather Research & Forecasting (WRF) Nucleus European Modelling Ocean (NEMO) models. Classical ML deep (DL) algorithms were compared analyzed their ability extract biogeochemical patterns Barents Sea. The proposed allows us forecast next 8 days based features preceding days. best R 2 metric achieved was 0.578 using LightGBM algorithm, confirming applicability developed solution map northern region even cases where unavailable period due insufficient illumination dense cloud cover.

Язык: Английский

---

Grain Crop Yield Prediction Using Machine Learning Based on UAV Remote Sensing: A Systematic Literature Review

Drones, Год журнала: 2024, Номер 8(10), С. 559 - 559

Опубликована: Окт. 8, 2024

Preharvest crop yield estimation is crucial for achieving food security and managing growth. Unmanned aerial vehicles (UAVs) can quickly accurately acquire field growth data are important mediums collecting agricultural remote sensing data. With the rapid development of machine learning, especially deep research on based UAV learning has achieved excellent results. This paper systematically reviews current through a search 76 articles, covering aspects such as grain crops studied, questions, collection, feature selection, optimal models, periods estimation. Through visual narrative analysis, conclusion covers all proposed questions. Wheat, corn, rice, soybeans main objects, mechanisms nitrogen fertilizer application, irrigation, variety diversity, gene diversity have received widespread attention. In modeling process, selection key to improving robustness accuracy model. Whether single modal features or multimodal research, multispectral images source information. The model may vary depending selected period but random forest convolutional neural networks still perform best in most cases. Finally, this study delves into challenges currently faced terms volume, optimization, determining period, algorithm limitations UAVs. Further needed areas augmentation, engineering, improvement, real-time future.

Язык: Английский

---