KOCHI: The Cochin University of Science and Technology (Cusat) has been granted a patent for developing a novel machine learning-based technology to extract rainfall information from atmospheric radar signals.
This method integrates distribution-based signal processing algorithms with advanced machine learning techniques to enhance the performance of the 205 MHz Stratosphere–Troposphere (ST) radar. Radar data often contain echoes from both clear-air and raindrops, making it difficult to accurately retrieve atmospheric parameters. To address this challenge, the research team developed a Hybrid Adaptive Bi-Gaussian Fitting Algorithm, capable of effectively separating rain and clear-air echoes within the troposphere below 5 km, even across varying rainfall intensities.
Furthermore, by applying the Gaussian Mixture Model (GMM), the technique demonstrated high efficiency in accurately distinguishing spectral components even under complex atmospheric conditions. The probabilistic nature of GMM enables precise identification of rain and air echoes.
The patented research is the outcome of a collaborative effort by Dhanya R, assistant professor, ToCH Institute of Science and Technology (former research scholar, Department of Electronics, Cusat), Anju Pradeep, professor, School of Engineering, Dr Abhilash S, director, Radar Centre, and Abhiram Nirmal, research scholar, Radar Centre.