Assessing Noise Effects on UAV Classification Accuracy With Deep Learning and FPGA Real-Time Processing: A Study Utilizing Radar Digital Twins
Bibliographic record
Abstract
The increasing use of Unmanned Aerial Vehicles (UAVs) highlights the need for robust classification systems. This study explores the impact of noise on UAV classification accuracy using radar-based deep learning. A noise-free dataset of range-Doppler maps (RDMs) is generated from digital twin simulations, and noisy datasets are created by adding Additive White Gaussian Noise (AWGN) across Signal-to-Noise Ratio (SNR) levels from -20 dB to 10 dB. A Deep Learning (DL) model, trained on a merged dataset, achieves an accuracy exceeding 98%. Performance evaluation shows accuracy dropping to 34% at -14 dB SNR, with Receiver Operating Characteristic (ROC) curves used for analysis. Inference at the software level is conducted using TensorFlow on a PC within the Vitis-AI Docker environment, achieving an accuracy of 96.39%. The quantized model, when deployed on the KR260 Field-Programmable Gate Array (FPGA), attains an accuracy of 94.73%. Although there is a slight drop in accuracy, the hardware implementation demonstrates impressive performance, supporting a real-time UAV classification system that is effective across different noise levels. Finally, the pre-trained model is deployed on an FPGA processing platform to evaluate its effectiveness for real-time classification. By leveraging the FPGA’s real-time processing capabilities, this approach satisfies the stringent speed and accuracy requirements necessary for UAV classification.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.001 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".