A study on Deep Learning based Soil Classification

Soil selection for good crop yielding plays a vital role and is a challenging task for farmers. Classification of soil requires analyses of features and characteristics of various soil of different regions. Soil can be classified based on its features like texture, particle size, and color by applying computer vision and image processing. However, deep learning-based soil classifications are producing better results. In Computer-based Soil classification, various Machine Learning algorithms such as K Nearest Neighbor, Quadratic Regression, Support Vector Machine (SVM), Naive Bayes, Decision Tree, Random Forest, Multi-Layer Perceptron (MLP), and Neural Networks are applied for soil classification by various researchers. The end-to-end process facilitated by deep learning largely reduces addiction to preprocessing techniques and spatial form designs. This paper is mainly focused on the study of various deep learning algorithms that are applied for soil classification. It talks about the varieties of features and databases used by researchers for soil classification using various deep learning techniques. Evaluation metrics are analyzed and shown for discrimination. Finally, a novel deep learning model for better soil classification is also proposed as our contribution to this study. This study helps the researchers understand various deep learning approaches implemented for soil classification. It also acts as a model to implement improved deep learning techniques in future research works.