Soil Classification Using Machine Learning, Deep Learning, and Computer Vision: A Review

Abstract

The goal of this work is to examine the primary soil variables that impact crop development, such as organic matter, vital plant nutrients, and micronutrients, and to determine the appropriate connection proportion among those qualities using ML and DL models. Agriculture relies heavily on the soil. Different kinds of dirt can be found in various locations. Each type of soil can have unique characteristics to develop crops. To cultivate crops in diverse soil types, we need to know which kind of soil is best. Accurate soil moisture and temperature forecast are beneficial to agriculture planting factors. According to the review articles, machine learning approaches, deep learning prediction models, and computer vision are helpful in this instance. The external temperature and humidity and soil moisture and temperature are used to train and test to anticipate soil moisture and temperature. A deep learning model was developed based on the extended short-term memory network (LSTM). Machine learning models such as k-nearest neighbor, SVM, and RF methods are utilized for soil classification. Several contributions of computer vision models in fields like planting, harvesting, enhanced weather analysis, weeding, and plant health detection and monitoring have been shown in the application of computer vision in soil categorization. This paper presents a review of soil classification and various challenges associated with it using machine learning, deep learning, and computer vision.