Abstract: MTF(modulation transfer function) serves as a critical parameter for imaging quality evaluation of infrared remote sensing satellites; however, its on-orbit detection faces two major challenges: insufficient thermal stability of calibration targets and complexity in data processing. This study proposed a direct on-orbit MTF detection method utilizing the synergistic three-bar and temperature-step targets. Through the target optimized design and fuzzy PID (proportional-integral-derivative) temperature control algorithm, the targets could achieve temperature uniformity (<0.5°C), stability (±0.6°C), and operational repeatability (>99%). By integrating Fourier optics theory and radiative transfer modeling with the computational principle of "object-image modulation separation", MTF values could be directly extracted at Nyquist frequency from remote sensing imagery. This approach could effectively eliminate errors caused by atmospheric interference, numerical differentiation, and parameter fitting inherent in conventional methods, providing high-precision technical support for quantitative applications of high-resolution infrared satellite systems.