Smart detection of plant nutrient deficiencies using machine learning and image fusion

Plant nutrient deficiencies are a continual challenge for enhancing global crop output and food security. Historically, manual inspections have been plagued by subjectivity, inefficiency, and restricted scalability, necessitating the development of enhanced detection algorithms. This research introduces a novel approach utilizing image processing and machine learning to enhance detection accuracy and practical applicability. This strategy promotes classification stability by integrating machine learning classifiers, including k-Nearest Neighbors, Artificial Neural Networks, Decision Trees, and Linear Discriminant Analysis, with fusion techniques such as Majority Voting and Mean Fusion. The experiments utilize Leave-One-Out Cross-Validation for model evaluation to address dataset variability and deliver thorough assessments. The study’s results indicate that the suggested system surpasses existing systems in accuracy, precision, recall, and F1 score, attaining an overall accuracy of 98.17%. The method is effective across various noise and resolution parameters, allowing for scalability in precision agriculture applications. This discovery not only enhances the diagnosis of plant nutrient deficiencies but also enables further investigations into real-time plant health monitoring.