Pair-Soil-Spectra: An Approach for NIRS-Based Soil Total Nitrogen Content Detection with Feature Metrics in Cases of Small Sample Sizes DOI
Yueting Wang, Chunjiang Zhao, Zhen Xing

et al.

Analytical Chemistry, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 19, 2024

Soil total nitrogen (STN) plays an important role in plant growth, and rapid nondestructive detection of STN content is essential for agricultural production. Near-infrared spectroscopy (NIRS) takes advantage the fast speed, low cost, nondestructiveness, it can be used detection. Typically, NIRS-based approaches require a large number samples model training. However, difficult to collect sufficient due various causes (e.g., time-varying state, high assay costs, etc.) practical application. To tackle this problem, feature metric approach introduced detect based on NIRS work, new (named Pair-Soil-Spectra) proposed mine fine-grained features by contrasting different soil sample pairs, which full particle heterogeneity penetration. For validation study, three datasets with collection sources are selected as research subjects, performance Pair-Soil-Spectra analyzed from perspectives. According results, has significantly improved models partial least-squares (PLS), Cubist, extreme learning machine (ELM), random forest (RF)) small cases. Of these, coefficient determination RF 0.13, 0.42, 0.10, root-mean-square prediction decreased 0.15, 0.52, 0.01 g/kg datasets, gained greatest improvement. Meanwhile, easily expanded cover other domains.

Language: Английский

A Novel Model for Soil Organic Matter and Total Nitrogen Detection Based on Visible/Shortwave Near-Infrared Spectroscopy DOI Creative Commons
Jiangtao Qi, Peng Cheng,

Junbo Zhou

et al.

Land, Journal Year: 2025, Volume and Issue: 14(2), P. 329 - 329

Published: Feb. 6, 2025

Soil organic matter (SOM) and total nitrogen (TN) are critical indicators for assessing soil fertility. Although laboratory chemical analysis methods can accurately measure their contents, these techniques time-consuming labor-intensive. Spectral technology, characterized by its high sensitivity convenience, has been increasingly integrated with machine learning algorithms nutrient monitoring. However, the process of spectral data remains complex requires further optimization simplicity efficiency to improve prediction accuracy. This study proposes a novel model enhance accuracy SOM TN predictions in northeast China’s black soil. Visible/Shortwave Near-Infrared Spectroscopy (Vis/SW-NIRS) within 350–1070 nm range were collected, preprocessed, dimensionality-reduced. The scores first nine principal components after partial least squares (PLS) dimensionality reduction selected as inputs, measured contents used outputs build back-propagation neural network (BPNN) model. results show that processed combination standard normal variate (SNV) multiple scattering correction (MSC) have best modeling performance. To stability this model, three named random search (RS), grid (GS), Bayesian (BO) introduced. demonstrate Vis/SW-NIRS provides reliable PLS-RS-BPNN achieving performance (R2 = 0.980 0.972, RMSE 1.004 0.006 TN, respectively). Compared traditional models such forests (RF), one-dimensional convolutional networks (1D-CNNs), extreme gradient boosting (XGBoost), proposed improves R2 0.164–0.344 predicting 0.257–0.314 respectively. These findings confirm potential technology effective tools prediction, offering valuable insights application sensing information.

Language: Английский

Citations

1
Soil organic matter content prediction in tobacco fields based on hyperspectral remote sensing and generative adversarial network data augmentation DOI
Yu Xia,

Xueying Cheng,

Xiao Hu

et al.

Computers and Electronics in Agriculture, Journal Year: 2025, Volume and Issue: 233, P. 110164 - 110164

Published: March 5, 2025

Language: Английский

Citations

0
Hyperspectral estimation of soil organic matter using improved spotted hyena optimizer and iteratively retained informative variables DOI
Hui Zhang, Yunbo Shen, Huanhuan Lv

et al.

Microchemical Journal, Journal Year: 2025, Volume and Issue: unknown, P. 113410 - 113410

Published: March 1, 2025

Language: Английский

Citations

0
Spatial Inversion of Soil Organic Carbon Content Based on Hyperspectral Data and Sentinel‐2 Images DOI

Xiaoyu Huang,

Xuemei Wang, Yanping Guo

et al.

Land Degradation and Development, Journal Year: 2025, Volume and Issue: unknown

Published: April 3, 2025

ABSTRACT Given that Sentinel‐2 (S2) multispectral images provide extensive spatial information and ground‐based hyperspectral data capture refined spectral characteristics, their integration can enhance both the comprehensiveness precision of surface acquisition. This study seeks to leverage these sources develop an optimized estimation model for accurately monitoring large‐scale soil organic carbon (SOC) content, thereby addressing current limitations in multi‐source fusion research. In this study, using mathematical transformation discrete wavelet transform process ground delta oasis Weigan Kuqa rivers Xinjiang, China, combination with S2 image, machine learning algorithms were employed construct models SOC content total variables characteristic variables, inversion oases was carried out. We found R ‐DWT‐H9 significantly correlation between ( p < 0.001). The accuracy constructed based on feature selected by SPA IRIV generally higher than variable models. IRIV‐RFR had highest stable capability. values 2 training validation sets 0.66 0.64, respectively. RMSE 1.5 g∙kg −1 , RPD > 1.4. interior oasis, mainly deficient (61.35%) or relatively (8.17%), while periphery it extremely (30.48%). Combine providing a reference evaluating fertility arid regions.

Language: Английский

Citations

0
Multi-spectral evaluation of total nitrogen, phosphorus and potassium content in soil using Vis-NIR spectroscopy based on a modified support vector machine with whale optimization algorithm DOI

Mochen Liu,

Yang Kuankuan,

Yinfa Yan

et al.

Soil and Tillage Research, Journal Year: 2025, Volume and Issue: 252, P. 106567 - 106567

Published: April 19, 2025

Language: Английский

Citations

0
Improving the accuracy of soil organic matter mapping in typical Planosol areas based on prior knowledge and probability hybrid model DOI

Deqiang Zang,

Yinghui Zhao, Chong Luo

et al.

Soil and Tillage Research, Journal Year: 2024, Volume and Issue: 246, P. 106358 - 106358

Published: Nov. 14, 2024

Language: Английский

Citations

2
Pair-Soil-Spectra: An Approach for NIRS-Based Soil Total Nitrogen Content Detection with Feature Metrics in Cases of Small Sample Sizes DOI
Yueting Wang, Chunjiang Zhao, Zhen Xing

et al.

Analytical Chemistry, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 19, 2024

Soil total nitrogen (STN) plays an important role in plant growth, and rapid nondestructive detection of STN content is essential for agricultural production. Near-infrared spectroscopy (NIRS) takes advantage the fast speed, low cost, nondestructiveness, it can be used detection. Typically, NIRS-based approaches require a large number samples model training. However, difficult to collect sufficient due various causes (e.g., time-varying state, high assay costs, etc.) practical application. To tackle this problem, feature metric approach introduced detect based on NIRS work, new (named Pair-Soil-Spectra) proposed mine fine-grained features by contrasting different soil sample pairs, which full particle heterogeneity penetration. For validation study, three datasets with collection sources are selected as research subjects, performance Pair-Soil-Spectra analyzed from perspectives. According results, has significantly improved models partial least-squares (PLS), Cubist, extreme learning machine (ELM), random forest (RF)) small cases. Of these, coefficient determination RF 0.13, 0.42, 0.10, root-mean-square prediction decreased 0.15, 0.52, 0.01 g/kg datasets, gained greatest improvement. Meanwhile, easily expanded cover other domains.

Language: Английский

Citations

0