The prerequisite for achieving precise mid-infrared (3 000~5 000 nm) radiation measurement is calibration. In recent years, the technology using correlated photons generated by Spontaneous Parametric Down-Conversion (SPDC) for calibration has attracted widespread attention as an absolute calibration method requiring no external standard transfer. This approach, based on an objective physical effect, enables self-calibration. However, limited by the absence of infrared single-photon detectors, research surveys indicate its primary applications have been confined to calibration within the visible-near-infrared spectral range. Up-conversion technology, as a novel infrared single-photon detection method, enables the detection of correlated photon signals in the mid-infrared band.This paper first introduces the principle and mathematical expression for determining channel detection efficiency through correlated photon calibration. Subsequently, an experimental optical path combining parametric down-conversion and up-conversion technologies was constructed, with detailed descriptions of the setup procedures, particularly emphasizing the principles of correlated photon focusing and beam combining. The experiment initially utilized a 775 nm laser to pump a Periodically Poled Lithium Niobate (PPLN) crystal, generating 1 005 nm/3 390 nm correlated photon pairs via SPDC. The 1 005 nm signal photons were routed through the trigger path, while the 3 390 nm idler photons were directed to the measurement path. To address the challenge of lacking mid-infrared single-photon detectors, the 3 390 nm idler photons in the measurement path were up-converted to 810 nm through sum-frequency generation with 1 064 nm pump light in another PPLN crystal. Silicon-based near-infrared single-photon detectors of the same model were used to detect the 1 005 nm and 810 nm photons separately. The outputs from both near-infrared single-photon detectors were connected to a Time-to-Digital Converter (TDC) for time-correlated coincidence measurements. Practical experimental issues including dead time, afterpulses, accidental coincidences, and differences in non-common optical paths were analyzed. A 3 390 nm laser was employed to measure the transmittance of non-common optical paths at 3 390 nm. Based on the modified expression for calculating channel efficiency in correlated photon calibration, the detection efficiency of the measurement path for 3 390 nm was determined by removing the influence of non-common optical path efficiency using the measured coincidence counts and trigger photon counts.The study further investigated the limitations on the area and solid angle of mid-infrared radiation in the measurement path. By comparing the acceptance areas and solid angles for mid-infrared radiation between the up-conversion section and the fiber-coupled collection section, it was concluded that the fiber-coupled collection component imposed the primary constraints on the solid angle and area of mid-infrared radiation. Combining the calibrated detection efficiency of the measurement path, an expression for calculating mid-infrared radiance from up-converted photon counts was derived.The calibrated system was used to measure the radiance of a cavity-type blackbody at 3 390 nm under temperatures of 125 ℃, 150 ℃, 175 ℃, and 200 ℃. Experimental results were compared with theoretical radiance values calculated using Planck's formula, revealing a maximum discrepancy of 4.3%. This demonstrates the potential application prospects of combining up-conversion and down-conversion technologies for absolute radiance measurement beyond 3 000 nm. Finally, possible causes for the observed discrepancies were analyzed, including potential uncalibrated blackbody emissivity and slight laser power fluctuations during the experiment. Future optimizations for the system were proposed, such as incorporating laser power stabilization modules, to further enhance measurement accuracy.