Enhancing Support Vector Machine Classification of Nutrient Deficiency in Rice Plants Through Particle Swarm Optimization-Based Feature Selection

Abstract

The research focuses on the classification of nutrient deficiencies in rice plant leaves using a combination of Support Vector Machine (SVM) and Particle Swarm Optimization (PSO) methods for feature selection. Image features are extracted using Histogram of Oriented Gradients (HOG), which is then optimized with PSO to select the most relevant features in the classification process. Indonesia is one of the largest rice producers in the world, with food security as a major issue that requires sustainable solutions, especially in the agricultural sector. The growth and yield of rice plants are highly dependent on the availability of nutrients such as Nitrogen (N), Phosphorus (P), and Potassium (K). However, traditional observation methods to detect nutrient deficiencies in plants become inefficient as the scale of production increases. The dataset used includes images of rice leaves showing nitrogen (N), phosphorus (P), and potassium (K) deficiencies. Experiments show that the SVM model optimized with PSO provides a classification accuracy of 83.19% and a runtime of 129.63 seconds with 1150 best feature combinations out of 2303 extracted features, which is higher accuracy and faster runtime than the model that does not use PSO. These results show that the integration of PSO in the feature selection process not only improves the accuracy of the model, but also reduces the required computation time. This research makes an important contribution to the development of an automated system for the classification of nutrient deficiencies in crops, which can be implemented in large farms or other agricultural fields.