Prediction of Soil Organic Matter with Deep Learning

Soil is the most important component of the ecosystem and the most significant characteristic of soil is its organic matter content, because organic matter undertakes many tasks by preventing soil moisture, absorption of water after rainfall, and good aeration by correcting bad textural properties and preventing soil erosion. Therefore, its recognition is critical, but the biggest problem is that determining soil organic matter with traditional methods is very laborious, expensive, and time-consuming. Accordingly, as in many different areas, computer vision methods can be used to determine soil organic matter. In this study, a new method based on deep learning has been proposed for the estimation of soil organic matter. In the study, firstly, images of 20 points where soil organic matter content was determined were taken with a special system. Then, a new segmentation method was applied to these images to separate the soil from the background and datasets were created. A new convolutional neural network was designed for organic matter estimation in these original datasets. In organic matter estimation, there is a difference of 0.01% between the proposed model and the value obtained in laboratory analysis. The proposed model is also compared with state-of-the-art deep learning models such as GoogleNet, ResNet, and MobileNet. In comparison, it has been seen that the proposed model is very successful in predicting organic matter with fewer parameters and in a shorter time, although it gives lower results with a slight difference in accuracy rates.