Automated Shape Classification Using SAR Imaging and Machine Learning

The proposed method involves generating milli-meter Wave Frequency Modulated Continuous Wave (mmWave FMCW) radar image data through MATLAB modeling, reconstructing images using SAR imaging technique, and classifying images that are cluttered with multiple object shapes such as triangles, circles, squares, donuts, T-shape, Polygon, Star, and Pentagon using a Random Forest classifier. The classifier’s performance is enhanced through hyper-parameter tuning and cross-validation. The model has high rate classification for T-shape Objects of 96.94% and minimum rate classification for Pentagon as 82.35% among all 9 type of object shapes. The overall model achieving high accuracy of 0.95%. The results demonstrate good accuracy in shape classification, validating the effectiveness of the integrated SAR and machine learning approach.