Abstract: This study investigates technical principles and advancements in the application of ground-based microwave radiometers for the observation of atmospheric temperature and humidity profiles, explores critical pathways that can enhance measurement accuracy and stability, and provides theoretical support for meteorological monitoring and climate research. Based on physical foundations including the thermal radiation theory, radiative transfer equations, and brightness temperature, this work systematically reviews microwave radiation measurement principles and calibration technology evolution as well as data quality control methods. By comparing performance parameters of typical domestic and international devices, this study focuses on analyzing the direction of retrieval algorithms optimization and the potential of integrating machine learning approaches. Ground-based microwave radiometers enable continuous detection of atmospheric temperature and humidity profiles within 0—10 km through multi-frequency observations (22—59 GHz). Liquid nitrogen cold calibration and tilt-curve calibration techniques improve brightness temperature accuracy up to 0.2—0.5 K. Neural network algorithms reduce the root mean square error (RMSE) of temperature inversion to 1.48℃, while physically constrained models decrease the mean absolute error (MAE) of high-altitude (＞8 km) temperature inversion by 0.19℃. However, challenges persist, including radio frequency interference in complex weather (10% error in L-band), insufficient long-term calibration stability (annual drift ＞0.2 K), and increased humidity inversion error under cloudy/rainy conditions (RMSE up to 25.21%). Ground-based microwave radiometers, enhanced by anti-interference hardware design and dynamic real-time calibration and machine learning-physical model fusion algorithms, can significantly improve the retrieval accuracy of atmospheric parameters. This research lays theoretical and technical foundations for their widespread application in meteorological monitoring and climate studies. Future efforts should prioritize breakthroughs in multi-frequency collaborative observations, nonlinear error correction, and extreme weather adaptability. Through in-depth exploration of environmental noise suppression, calibration error tracking, and retrieval robustness enhancement, this work provides a critical support for advancing atmospheric detection technologies under global climate change, facilitating the development of high-precision, all-weather atmospheric profile observation systems.