A Hybrid Model for Missile Detection Using YOLOv8 and Machine Learning Enhancement

Missiles often overstate their capabilities, yet they are both intricate and simple machines. They are not a modern phenomenon, but numerous countries operate them nowadays due to the spread of the technical expertise and equipment needed to develop them. So, missiles represent a serious threat if they aren't terminated in time. In military operations, the object detection task plays a crucial role, which is attributed to its ability to identify and classify objects in generated images. Recent studies in the field of missile detection have also lacked research that merges deep learning (DL) with machine learning (ML) methods, especially the YOLOv8 and Random Forest (RF)/XGBoost models, which present a unique opportunity to assess the effects of this combination on accuracy and efficiency. The hybrid models involved training with a large dataset, exceeding 10000 images collected and annotated from the Rodoflow site. The YOLOv8 algorithm was used to detect and identify missiles, and then the enhancement layer was applied. The model was trained, validated, and tested using Python. To ensure the reliability of the results and the generalizability of the model, performance was evaluated using independent test data that wasn't used during the training phase, to verify the model's ability to work efficiently on previously unseen data. The model achieved high accuracy, including 80.8% mAP@50 with 80.8% precision for YOLOv8 and 77.1% mAP with 93.9% precision for RF and 77.0% mAP with 90.9% precision for XGBoost. The paper noted that the hybrid model with RF produced a significant result compared to the model with XG.