Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{197697,
        author = {Mr. Abdul Rais and Ghina Fatima and Sania Tabassum and Urooj Fatima Quadri and Mohammed Abdul Mudassir Ansari},
        title = {A Novel Approach for Disaster Victim Detection Under Debris Environments Using Decision Tree Algorithms with Deep Learning Features},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {7281-7294},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=197697},
        abstract = {Rapid victim identification in collapsed building environments remains one of the most time-critical challenges in Urban Search and Rescue (USAR) operations. This paper proposes a novel hybrid deep learning and machine learning framework for Human Victim Detection (HVD) under debris environments. A custom RGB image dataset comprising five class labels — head, hand, leg, upper body, and no body — was created and augmented to improve diversity and size. Transfer Learning on the ResNet-50 architecture is employed to extract rich, discriminative feature representations from the dataset. A J48 Decision Tree algorithm is subsequently applied for feature selection, discarding irrelevant features before classification. Multiple machine learning classifiers — Support Vector Machine (SVM), Random Forest, Multi-Layer Perceptron (MLP), Naive Bayes, and the proposed extension XGBoost — are then evaluated on the selected features. Experimental results demonstrate that the XGBoost extension achieves the highest classification accuracy of 98%, followed by Random Forest at 96% and SVM at 95%, with inference times well within real-time operational constraints. The results confirm that integrating deep learning-based feature extraction with machine learning classification significantly outperforms standalone approaches, offering a reliable and computationally efficient solution for automated victim detection in disaster scenarios.},
        keywords = {Disaster Victim Detection; Transfer Learning; ResNet-50; Decision Tree; Random Forest; XGBoost; SVM; Machine Learning; Deep Learning; Urban Search and Rescue; Feature Selection; Debris Environments},
        month = {April},
        }