Lightweight Machine Learning Models for Drone Detection Using Acoustic and Optical Features

The widespread deployment of drones has triggered major concerns over privacy and security, creating a demand for robust anti-drone systems (ADS). A critical component of ADS is the detection unit/model, which identifies drones in an unauthorized area. Recently, various statistical and machine/deep learning methods have been developed for drone detection units. Statistical methods are traditionally applied which often suffer from uncertain thresholds under varying noise distributions. While deep learning-based methods are highly popular, they frequently face challenges related to high computational complexity. This study explores the potential of low-complexity machine learning (LCML) models, including logistic regression model (LRM), support vector machines (SVM), and random forest algorithm (RFA) for drone detection using acoustic and optical features. For optical-based detection, histogram of oriented gradients (HOG) features extracted from images are used, whereas for acoustic-based detection HOG features are derived from log-mel spectrograms of drone acoustic signals. The LCML models are assessed using various performance metrics for binary classification, with RFA demonstrating the best results that achieves 87.5% accuracy with optical features and 89% accuracy with acoustic features. In addition, it outperforms SVM and LRM irrespective of the input feature. However, LRM exhibits the lowest training and testing complexity, making it a preferable choice where limited computational resources are available. These findings suggest that RFA is the most promising LCML model for real-time detection unit for an ADS that offers a balance between accuracy and inference complexity.