Collaborative Projects

Overview:

Passive microwave remote sensing plays a crucial role in environmental monitoring, particularly in applications such as precision agriculture, water management, and soil moisture estimation. Conventional satellite-based microwave radiometers offer coarse spatial resolutions, limiting their usefulness in precision agriculture scenarios. To bridge this gap, we have designed and developed a portable, dual-polarized L-band microwave radiometer that operates from small Unmanned Aircraft Systems (UAS), offering high-resolution measurements of surface brightness temperature (TB).

Key Contributions:

• Innovative Radiometer Design:

  • Developed a dual-polarized (H-pol and V-pol) antenna paired with a software-defined radio (SDR)-based receiver, enabling flexible, high-resolution measurements from small aerial platforms.

  • The system operates within the protected L-band frequency (1400–1427 MHz), ensuring minimal interference and optimal sensitivity for soil moisture estimation.

• Advanced Calibration and Stability:

  • Implemented a rigorous two-step calibration process, including internal calibration using precise temperature references (liquid nitrogen, dry ice, ambient hot source) and external calibration using sky and blackbody measurements.

  • Achieved remarkable measurement stability, with low standard deviation in brightness temperature readings (0.39 K for blackbody measurements and 0.95 K for sky measurements).

• Robust Onboard Data Processing:

  • Integrated advanced signal processing techniques capable of removing Radio Frequency Interference (RFI) from raw IQ data samples.

  • Demonstrated onboard RFI detection and mitigation capabilities, significantly enhancing measurement accuracy and data reliability during real-time aerial missions.

• High-Resolution Surface Mapping:

  • Successfully tested the radiometer mounted on a custom-made octa-copter UAS over a mixed land-water environment.

  • Generated accurate, high-resolution brightness temperature maps clearly delineating land and water boundaries, enabling detailed monitoring of surface features and moisture conditions.

• Flexible and Cost-Effective Deployment:

  • Designed a compact, lightweight, and cost-effective radiometer system, optimized for small UAS platforms, with a total payload weight of approximately 9 kg (20 lbs).

  • Facilitated near-real-time data collection and processing capabilities, suitable for practical, on-site field deployment in agricultural and environmental monitoring scenarios.

Results and Impact:

• Superior Spatial Resolution:

  • Achieved a significantly higher spatial resolution (approximately 22m x 15m ground footprint) compared to traditional satellite-based radiometers, effectively supporting precision agriculture practices at the subfield scale.

• Real-Time Environmental Monitoring:

  • Provided reliable real-time surface brightness temperature measurements, critical for soil moisture estimation, irrigation scheduling, and drought monitoring, offering valuable decision-support data for agricultural practitioners.

• Advancement in RFI Mitigation Techniques:

  • Demonstrated the practical viability of employing onboard, real-time RFI detection and mitigation strategies, significantly enhancing the accuracy and reliability of microwave remote sensing data.

Future Directions:

• Further miniaturization of the radiometer payload to extend UAS flight duration and enhance coverage for larger agricultural fields.

• Development of real-time data analytics and adaptive measurement techniques to dynamically respond to changing environmental conditions.

• Implementation of the system within broader testbeds for research in active-passive microwave remote sensing coexistence, contributing further to the field of spectrum-efficient sensing technologies.

This UAS-based SDR dual-polarized L-band microwave radiometer represents a significant step forward in high-resolution, cost-effective, and accurate passive remote sensing, with transformative implications for environmental management and precision agriculture applications.

This project presents a novel deep learning framework specifically designed for efficient and real-time bearing fault diagnosis in rotary machinery. The method introduces parameterized learnable filters—specifically Gaussian and Sinc filters—integrated directly into a convolutional neural network (CNN) architecture, enabling the model to process raw multi-channel vibration data without the need for computationally intensive time-frequency transformations. Unlike traditional CNN methods that rely on extensive preprocessing and large datasets, our approach significantly reduces computational latency, achieves state-of-the-art accuracy (up to 99.5%), and maintains high performance even with limited training samples. Moreover, the use of structured learnable filters provides clear physical interpretations, highlighting which frequency bands are critical for fault detection. Validated extensively on established bearing fault datasets, this method demonstrates superior classification performance, efficiency, and interpretability, making it highly suitable for real-time condition monitoring and predictive maintenance applications.