Ensemble Learning for UAV Detection: Developing a Multi-Class Multimodal Dataset

Abstract

Unmanned aerial vehicles (UAVs) are a growing threat to public safety if used maliciously. In this study, we present our multimodal data set containing image, audio, and radio frequency (RF) data, which can serve as a valuable resource for researchers and developers in the field of UAV detection. We present a multiclass multimodal ensemble approach to address the need to improve UAV identification and detection. Our approach is novel as we integrated multiple deep-learning classifiers into a single ensemble classifier. We evaluate the performance of our proposed solution with a hard-voting model and a soft-voted model to evaluate the effectiveness of the proposed solution. Overall, our ensemble approach performed better than the single-modality classifier and when combined, could mitigate the low accuracy of the RF (CNN) accuracy score of 67%. This study has shown how effective ensemble approaches can be used to mitigate limitations when predicting UAV based on multimodal signatures.