Controlled image-based prediction of soil nutrient concentrations (NPK) using machine and deep learning

Soil nutrient analysis is essential for sustainable agriculture and precision farming. Traditional chemical tests for nitrogen, phosphorus, and potassium are reliable but costly, time-consuming, and often unavailable in remote areas. The objective of this study is to develop a novel approach that combines laboratory chemical extraction, controlled image acquisition, image processing, machine learning, and deep learning to classify soil nitrogen, phosphorus, and potassium levels. 50 independent soil samples were analyzed, chemically extracted, and classified as Low, Medium, or High for each nutrient. A specially designed black-box prototype was developed to capture high-quality images of soil extract samples under consistent lighting conditions. The images were used to extract color and texture information, which was evaluated using multiple classical machine-learning algorithms, such as K-nearest neighbours, support vector machines, logistic regression, random forest, extra trees, gradient boosting, and Gaussian naive Bayes. A convolutional neural network-based model was also used for comparison. To minimize data leakage, all validation procedures were performed at the physical-sample level, and augmented images were restricted to the training data. The best models obtained using nested cross-validation for nitrogen, phosphorus, and potassium were KNN, Random Forest, and Logistic Regression, respectively, with accuracies of 60%, 82%, and 48%. The nitrogen model correctly identified the High category, and the phosphorus model correctly identified the dominant Low category each time. For potassium, Logistic Regression made predictions of all three classes and outperformed the majority class baseline. The Classical machine-learning models slightly outperformed the compact CNN for nitrogen and potassium, while both achieved the same accuracy for phosphorus; however, overall performance remained limited by the small and imbalanced dataset. The findings suggest that the proposed framework could contribute to image-assisted decision-making in smart farming and support improved soil management practices, particularly in regions where laboratory testing facilities are limited.