Rapid and Accurate Prediction of Soil Texture Using an Image-Based Deep Learning Autoencoder Convolutional Neural Network Random Forest (DLAC-CNN-RF) Algorithm

Soil determines the degree of water infiltration, crop nutrient absorption, and germination, which in turn affects yield quality. For efficient planting agricultural products, accurate identification soil texture is necessary. This study proposed a flexible smartphone-based machine vision system using deep learning autoencoder convolutional neural network random forest (DLAC-CNN-RF) model for identification. Different image features (color, particle, texture) were extracted randomly combined to predict sand, clay, silt content via RF DLAC-CNN-RF algorithms. The results show that has good performance. When full extracted, very high prediction accuracy sand (R2 = 0.99), clay 0.98), 0.98) was realized, higher than those frequently obtained by KNN VGG16-RF models. possible mechanism further discussed. Finally, graphical user interface designed used accurately types. investigation showed could be promising solution costly time-consuming laboratory methods.