Machine Learning-Based Monostatic Microwave Radar for Building Material Classification

A Microwave nondestructive testing and evaluation (MNDT&E) method combining a broadband continuous wave (CW) monostatic radar and machine learning (ML) algorithms is proposed for building materials classification. A low-power and compact system using commercial solutions is considered. The data collected, i.e. complex reflection coefficient S11, are pre-processed then used as inputs to train support vector machine (SVM) and decision tree classifier (DCT) algorithms. The trained ML models achieved accuracy of 97.88% for DCT and 88.60% for SVM confirming the interest of the method for contactless building material classification.