A sophisticated radar-based object classification system that distinguishes between UAVs (drones), birds, RC aircraft, and mixed scenarios using advanced signal processing techniques and deep learning.

Features

Advanced Signal Processing: Implementation of Successive Variational Mode Decomposition (SVMD) for superior signal analysis

Cross-term-free Spectrograms: Enhanced time-frequency representations for better classification accuracy

Transfer Learning: Utilizes pre-trained SqueezeNet for efficient feature extraction

Multi-class Classification: Distinguishes between drones, birds, RC planes, and mixed scenarios

Automated Pipeline: End-to-end processing from raw radar data to classification result

Methodology

Reference Video

For a better understanding of the overall approach and implementation, please refer to the following video:

👉 Video Tutorial

Note: The final implementation in this repository may not exactly match the one demonstrated in the video. However, the project is largely inspired by the concepts and methodology presented there, and the video can be used as a helpful resource for understanding the underlying ideas and workflow.

Signal Processing Pipeline

Preprocessing: Downsampling, resampling, and low-pass filtering

SVMD Decomposition: Successive decomposition into Intrinsic Mode Functions (IMFs)

Spectrogram Generation: Cross-term-free STFT computation

Feature Extraction: Deep CNN features using SqueezeNet

Classification: Multi-class object identification

Key Algorithms

SVMD (Successive Variational Mode Decomposition): Advanced signal decomposition technique

VMD (Variational Mode Decomposition): Standard mode decomposition for comparison

Transfer Learning: Pre-trained CNN feature extraction

Prerequisites

MATLAB R2020b or later (using 2023b)

Signal Processing Toolbox

Deep Learning Toolbox

Image Processing Toolbox

Installation

git clone https://github.com/diptiman-mohanta/Radar-Based-UAV-Classification.git
cd Radar-Based-UAV-Classification

Citation

If you use this work in your research, please cite:

@misc{radar_uav_classification,
  title={Radar-Based UAV Classification using SVMD, Spectogram and Deep Learning},
  author={Diptiman Mohanta and Akash S R and Shekh Sharfraj and Krishna Jyoti Panda and Arpita Pradhan and Jyotirmayee Patnaik},
  year={2025},
  url={https://github.com/diptiman-mohanta/Radar-Based-UAV-Classification.git}
}

Dataset

@data{1x2q-8v62-22,
doi = {10.21227/1x2q-8v62},
url = {https://dx.doi.org/10.21227/1x2q-8v62},
author = {Harish Chandra Kumawat and Mainak Chakraborty and A. Arockia Bazil Raj and Sunita Vikrant Dhavale},
publisher = {IEEE Dataport},
title = {DIAT-µSAT: micro-Doppler Signature Dataset of Small Unmanned Aerial Vehicle (SUAV)},
year = {2022} }

to use the dataset i have used mail to the authors to provide the .mat files you can get the email id in this [https://ieee-dataport.org/documents/diat-msat-micro-doppler-signature-dataset-small-unmanned-aerial-vehicle-suav]