URES-Net: A Unified Region Extraction and Hybrid Deep Learning Framework for Intelligent Soil Quality Assessment

Soil quality assessment is a fundamental part of
sustainable agriculture and precision farming, and it impacts
directly on the productivity of crops and environmental
management. Traditional techniques used to analyze soil in the
laboratory are accurate but time-consuming, expensive, and also
not feasible for large-scale or real-time monitoring, and current
methods based on machine learning find it difficult to handle
noisy data, localize the region poorly and also lack generalization.
To overcome these challenges, in this research, URES-Net, a
Unified Region Extraction and Hybrid Deep Learning
framework for intelligent soil quality assessment, is proposed.
The main goal is to correctly classify the soil quality levels using
region-aware feature extraction with the support of hybrid deep
learning architectures. URES-Net integrates the region extraction
module in order to extract informative soil texture and color
patterns, and a hybrid CNN-BiLSTM network with the ability to
extract both spatial and contextual dependencies. Experiments
performed on benchmark soil image datasets show that the
proposed framework provides with 96.8% accuracy, F1-score of
96.3%, and precision of 97.1% which is considerably better than
five state-of-the-art models by margins of 3.2% to 8.5%. The
results validate that unified region extraction and hybrid learning
methods largely improve the prediction of soil quality, and
URES-Net is a powerful and scalable solution for intelligent
agricultural decision support systems.
Keywords: Soil Quality Assessment, Deep Learning, Region
Extraction, Hybrid Neural Networks, Precision Agriculture,
Computer Vision