Trapped human victims Machine Learning detection Model Using Recursive Feature Elimination with SVM Classification Algorithm

Building collapses, irrespective of structure type or location, result in global tragedies. Rapid rescue efforts are essential, especially within the critical three-day window for trapped victims. While disaster relief techniques incorporating hardware, signal processing, and machine learning have evolved, further enhancements are needed, particularly in regions like Africa where collapses are prevalent. This study focuses on non-line-of-sight (NLOS) human detection, aiming to improve classification and localization in the aftermath of building collapses. After analyzing 17,408 instances, recursive feature elimination (RFE) is employed for dimensionality reduction. The reduced dataset undergoes support Vector machine (SVM) classification, achieving an impressive 86.3% accuracy. Comparative analyses against cutting-edge methods underscore the efficacy of the proposed approach. Contributing to both theoretical foundations and practical applications in search and rescue (SAR) operations, this research advocates for the integration of feature selection and extraction techniques in future studies. Finally, this work advances NLOS human detection, laying a foundation for future research by emphasizing RFE iteration refinement, exploring complementary techniques, and incorporating additional classifiers for comprehensive performance enhancement. The findings hold significant implications for SAR teams, offering valuable insights into effective strategies for prediction, detection, and localization of victims following structural collapses.