| Classification of small radar cross section targets with Convolutional Neural Networks (CNNs) |
| Paper ID : 1086-IUGRC5 (R1) |
| Authors: |
|
mohamed ali mohamed * Military Technical Colledge |
| Abstract: |
| In the recent years, drones' usages had been exploded widely in many useful applications as they are of simple use and cost-effective solutions for many daily life problems. On the other side, drones facilitate malicious work such as smuggling and other things that are forbidden by the law which oblige researchers to innovate in the field of drone classification and recognition. In this work, a CNN algorithm is introduced for drone classification based pre-processed data coming from a LFCW radar. The pre-processing stage is using the EMD as a denoising algorithm. The CNN is trained based on a synthesized data set emulating raw data coming from the radar. This work achieves a classification accuracy of 90% for SNR above 2 dB allowing the system to in security and defence applications to discriminate between drones and other entities. So, nowadays, classification of drones is one of the most important objectives for the researchers to decrease crimes made by these drones. Classification of drones, nowadays, is made using radars due to it is working without respect to the weather, so the radars must be trained for this work. The best way to train the radars is Artificial Intelligence specially with CNNs Deep Learning method which select the target features itself without needing to human interference. Also, as known that the RCS of drones is comparable with birds and this leads researchers to create much more accurate algorithms to have the best classification accuracy. |
| Keywords: |
| Convolutional Neural Networks, EMD, Radar Classification. |
| Status : Paper Accepted (Oral Presentation) |
