Soil pH is one of the major factors to be considered before doing any cultivation. Farmers always tested their soil pH either in soil pH laboratory, soil pH color chart or sometimes with the help of an expert. But these methods need time, labor and expertness. In this paper, a digital Smartphone image-based method is presented which predicts the soil pH in a simple and accurate way. Soil images are captured with the help of Redmi 3S prime Smartphone and store all the images as soil dataset.

more »

... images are processed through the different steps of digital image processing including soil image enhancement, soil image segmentation, and soil image feature extraction. During the feature extraction, Hue, Saturation and Value of the soil image are calculated and store Saturation and Hue plus Saturation as an index for the feature vector of the soil images. Prediction of soil pH is done with the help of Linear Regression, Artificial Neural Network, and KNN Regression. The coefficient of the linear regression is 0.859 for the Saturation feature of the soil image. Again, the coefficient of linear regression is 0.823 for Hue plus Saturation. The regression coefficient for ANN is 0.94064 with Levenberg-Marquardt algorithm and 0.92932 with Scaled Conjugate Gradient Backpropagation Algorithm. The regression coefficient of KNN is 0.89326 for K=5 with an RMSE value 0.1311. It is found that ANN always gives a better result as compare to another one.