Abstract:Soil texture is one of the most important indicators of soil physical properties, which has traditionally been measured through laborious procedures. Approaches utilizing visible near-infrared spectroscopy, with their advantages in efficiency, eco-friendliness and non-destruction, are emerging as potent alternatives. Nevertheless, these approaches often suffer from limitations in classification accuracy, and the substantial impact of spectral preprocessing, model integration, and sample matrix effect is commonly disregarded. Here a novel 11-class soil texture classification strategy that address this challenge by combining Multiplicative Scatter Correction (MSC) with Residual Network (ResNet) models was presented, resulting in exceptional classification accuracy. Utilizing the LUCAS dataset, collected by the Land Use and Cover Area frame Statistical Survey project, we thoroughly evaluated eight spectral preprocessing methods. Our findings underscored the superior performance of MSC in reducing spatial complexity within spectral data, showcasing its crucial role in enhancing model precision. Through comparisons of three 1D CNN models and two ResNet models integrated with MSC, we established the superior performance of the MSC-incorporated ResNet model, achieving an overall accuracy of 98.97 % and five soil textures even reached 100.00 %. The ResNet model demonstrated a marked superiority in classifying datasets with similar features, as observed by the confusion matrix analysis. Moreover, we investigated the potential benefit of pre-categorization based on land cover type of the soil samples in enhancing the accuracy of soil texture classification models, achieving overall classification accuracies exceeding 99.39 % for woodland, grassland, and farmland with the 2-layer ResNet model. The proposed work provides a pioneering and efficient strategy for rapid and precise soil texture identification via visible near-infrared spectroscopy, demonstrating unparalleled accuracy compared to existing methods, thus significantly enhancing the practical application prospects in soil, agricultural and environmental science.

Multi-fidelity fusion for soil classification via LSTM and multi-head self-attention CNN model

A Multifidelity Neural Network (MFNN) for Constitutive Modeling of Complex Soil Behaviors

Multiscale 3-D-2-D Mixed CNN and Lightweight Attention-Free Transformer for Hyperspectral and LiDAR Classification

Multi-modal Fusion Deep Learning Model for Excavated Soil Heterogeneous Data with Efficient Classification

Automatic soil classification method from CPTU data based on convolutional neural networks

Simultaneous Prediction of Soil Properties Using Multi_CNN Model

A classification framework with multi-level fusion of object-based analysis and convolutional neural network: a case study for land use classification in mining areas

Multi-Scale Spatial Attention-Based Multi-Channel 2D Convolutional Network for Soil Property Prediction

Multibranch Feature Fusion Network with Self- and Cross-Guided Attention for Hyperspectral and LiDAR Classification

A Mutual Guidance Attention-Based Multi-Level Fusion Network for Hyperspectral and LiDAR Classification

Spectral Fusion Modeling for Soil Organic Carbon by a Parallel Input-Convolutional Neural Network

Soil carbon content prediction using multi-source data feature fusion of deep learning based on spectral and hyperspectral images

Cross Attention-Based Multi-Scale Convolutional Fusion Network for Hyperspectral and LiDAR Joint Classification

Multi-layer feature fusion for hyperspectral image classification

Lithological Mapping Based on Fully Convolutional Network and Multi-Source Geological Data

A multiple model machine learning approach for soil classification from cone penetration test data

Accurate landslide identification by multisource data fusion analysis with improved feature extraction backbone network

Enhanced crop classification through integrated optical and SAR data: a deep learning approach for multi-source image fusion

A Novel Local-Global Dependency Deep Learning Model for Soil Mapping

Multilayer Feature Fusion Network With Spatial Attention and Gated Mechanism for Remote Sensing Scene Classification

Enhancing soil texture classification with multivariate scattering correction and residual neural networks using visible near-infrared spectra