Now 6398 visitors
Today:0 Yesterday:0
Total: 152780
216S 51P 5R
2026-10-22, Week 43 |
| Member Login |
| TACT Journal Homepage |
| Call for Paper |
| Committee |
| Find My Paper |
| Author Homepage |
| Statistics & History |
| Welcome Message |
| Paper Procedure |
| FAQ |
| Paper Archives |
| Outstanding Papers |
| Author Homepage |
| - Paper Procedure |
| - Journal Procedure |
| - Presentation Tips |
|
| Program & Proceedings |
| Program with Papers |
| Plenary Session |
| Tutorial Session |
|
| Presentation Platform |
| Hotel & Travel Info |
| About Korea |
| Accommodation |
| Transportation |
| VISA |
| Other Infomation |
|
| Photo Gallery |
| Scheduler Login |
| Seminar |
| Archives Login |
|
| Sponsors |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| Save Q&A |
 |
* Edit or answer any Q&A by selecting Q&A number Hyper Link below + Write button (Save)
|
ICACT20260373 Question.1
Questioner: qga0809@chungbuk.ac.kr 2026-03-01 오후 7:08:54 |
ICACT20260373 Answer.1
Answer by Auhor 758743342@qq.com 2026-03-01 오후 7:08:54
Chrome  Click!! |
| You validated your model on simulated and drone-measured data. How much does the recognition accuracy drop when the drones flight altitude goes beyond the 4 to 15 meter range used in your experiments? |
Thank you for your insightful question. In our current study, the drone-mounted experiments were conducted within a hover altitude range of 4 to 15 meters, which was selected to reflect typical operating conditions for low-altitude terrain sensing. While we have not yet systematically evaluated the models performance outside this range, we can infer potential behavior based on the design of our approach. The GAN-based model is trained on power spectral features extracted from echo signals, which are influenced by altitude primarily through changes in signal attenuation and multipath effects. Within the tested range, the model demonstrated stable and high classification accuracy (over 98%), suggesting robustness to altitude variations within this interval. If the flight altitude exceeds 15 meters, we anticipate a gradual decline in recognition accuracy, mainly due to the reduced signal-to-noise ratio (SNR) and increased beam footprint, which may smooth out fine-grained terrain features. Conversely, at altitudes below 4 meters, the signal may become overly sensitive to small-scale surface irregularities, potentially introducing noise and reducing generalization. That said, thanks to the domain adaptation capability of GANs and the use of simulated data covering a wider range of altitudes (4–15 m in 1 m increments) during training, the model may still generalize reasonably beyond the tested range. Nevertheless, further experiments at extended altitudes would be valuable to quantify the exact performance drop and to refine the model for broader operational envelopes. We are currently planning such experiments as part of future work to validate the models altitude generalization limits. |
ICACT20260373 Question.2
Questioner: abilaaniloshan@gmail.com 2026-03-01 오후 7:10:13 |
ICACT20260373 Answer.2
Answer by Auhor 758743342@qq.com 2026-03-01 오후 7:10:13
Chrome Click!! |
| How does your GAN-based approach specifically address the small-sample problem better than conventional CNN-based classifiers? |
Thank you for your question. In our GAN-based approach, the small-sample problem is addressed through a data augmentation mechanism inherent to the adversarial training process. Unlike conventional CNN-based classifiers, which rely solely on the original training set and are prone to overfitting when data is limited, our method leverages the generator network to synthesize diverse and realistic power spectral signals that expand the effective training distribution. Specifically, the generator learns to produce terrain-like echo spectra from random noise, guided by the discriminator’s feedback. This process not only increases the quantity of training samples but also enriches their feature diversity, helping the model generalize better to unseen data. Moreover, the dual-probe discriminator structure in our framework simultaneously performs discrimination and classification, enabling the model to learn shared feature representations that are robust to data scarcity. In contrast, conventional CNNs typically require large labeled datasets to learn generalizable features and often struggle with limited samples, even when data augmentation techniques like rotation or noise injection are applied. Our GAN-based framework, by contrast, generates semantically meaningful and domain-relevant samples, which leads to improved classification performance and reduced overfitting, as evidenced by the high accuracy (over 98%) achieved across all terrain categories in our experiments. |
| |
