Answers are grounded in this paper and cite their sources.
The identical nature of range-Doppler/micro-Doppler (r-D/m-D) signatures of small unmanned aerial vehicle (sUAV) invite the involvement of artificial intelligence (AI) for their detection/classification. Designing an appropriate machine learning (ML) technique to have an AI classifier having less training/testing time, less computational cost, and need of less hardware resources; but at increased accuracy, is an immediate requirement, to handle the threads of today’s sUAV targets. To address this research goal, an in-house X-band (10 GHz) continuous wave (CW) radar is built and used to collect the r-D/m-D time-Frequency (T-F) signatures of six different sUAV targets. A dataset "DIAT-μSAT" consists of totally a 4860 r-D/m-D T-F images is formulated/used to design/propose a hyper-parameters-tuned randomised principal component analysis (R-PCA) + support vector machine (SVM) classification technique, which is optimised at its input image resolution, training time, hyper parameters, feature extractor (kernel), and classification accuracy. The performance of proposed classifier is compared with other relevant state-of-the-art (SOTA) ML techniques, validated with several open-filed experiments conducted during different atmospheric conditions and the respective results are reported/analysed. The overall classification accuracy of 98% is achieved with 8.11e-4s test time, 12.68GB RAM, 107.72GB disk space, and 1.24 Mb memory size hardware platform, which is more suitable for on-the-fly (real-time) sUAV targets classification applications.
No takes yet. Share an insight, caveat, or question.
Hasan et al. (2023) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: