Digital Soil Texture Classification Using Machine Learning Approaches

The texture of the soil is an important factor to consider during cultivation. The water transmission property is being regulated by the texture of the soil. To determine sand, silt and clay’s percentage present in a soil sample, a conventional laboratory method is used, which consumes more time. Digitization in agriculture has given a new direction of innovative research in agriculture domain. In this paper, based on image processing an efficient model has been developed for soil texture classification. Eight different image preprocessing techniques were used for the image enhancement. Out of that, the linear contrast adjustment found to be best in image enhancement. A feature vector was calculated by extracting six different features from the enhanced image. The feature vector of an image is input to the machine learning classifier. The various classifiers used in this research work are SVM, KNN, ANN and PNN. The accuracy of the classifiers was SVM (0.98), KNN (0.89), ANN (0.89) and PNN (0.86). From the result, it is found SVM model has higher rate in classification of soil.