Multi-Feature Fusion for Soil Image Feature Extraction and Classification Using Machine Learning

Classification of the soil is one of the essential agricultural tasks for effective yield. Generally, soil classification has been done using conventional laboratory methods, which are time-consuming and high-cost. Computer-aided image analysis and artificial intelligence play a prominent role in classifying objects. This paper uses soil images to classify the soil type to overcome the long computational time and high cost. The texture of the soil is a critical factor in deciding the soil type. In this research, we proposed a multi-feature fusion in soil image feature extraction and classification using machine learning (ML). The soil samples were gathered from various regions of Andhra Pradesh, India, and images were captured by following certain criteria. The texture features: Gray Level Co-occurrence Matrix, Tamura, Gabor filter banks, and HSV texture features are fused to identify the soil type. The fused features are used in classifying the soil images using machine learning (ML) algorithms. The performance analyses in classifying the soil type using the Gaussian-kernel Support Vector Machine are better when compared with other ML algorithms.