Localization of Dielectric Anomalies with Multi-level Outlier Detection through Membership Function and Ensemble Classification Framework

This research presents an innovative method for real-time detection of dielectric anomalies, with a primary focus on evaluating apple quality and ripeness using dielectric tomography. The study involves the development of an advanced tomography system within an anechoic chamber, harnessing electromagnetic wave technology and sophisticated antenna systems for data acquisition. The proposed framework encompasses critical stages, including data collection, range bounds computation, threshold determination, class membership assignment, and ensemble classification. By seamlessly integrating statistical methods, density-based clustering, and ensemble learning, this approach significantly enhances precision and reliability in anomaly detection. The integration of available statistical methods, density-based clustering, and ensemble learning may demand substantial computational resources, limiting the scalability and real-time applicability of the proposed framework. Empirical results demonstrate the superior performance of the method, with an accuracy rate of 98.9%, precision of 0.989, F-measure of 0.989, dielectric anomaly recall rate of 0.99, and a low error rate of 0.18. Overall, this research introduces an advanced approach with the potential to revolutionize apple quality assessment and industrial processes across various sectors.