AC air arcs are generated in medium-voltage (MV) power systems under the effect of harsh weather conditions, equipment aging, and high penetration of distributed generation, threatening equipment and public safety. The arc current and temperature are low due to the wide application of arc suppression devices. In this scenario, the MV AC air arc does not satisfy the local thermodynamic equilibrium (LTE) condition. In addition, the repeated arcing and extinguishing processes further complicate the arc discharge mechanism, which bring challenges in the modeling and detection of MV AC air arcs. Experimental methods are a direct and efficient approach to determine the properties of arc plasmas. In this study, a dual-wavelength moiré deflection diagnostic system was established to determine the time evolution of the particle density and radial distribution of the temperature in an MV AC air arc without relying on the LTE assumption. The electron number density and heavy particle number density change transiently during the arc discharge process and change gradient along the radial direction. The heavy particle temperature and electron temperature were then calculated based on the measured particle number density. During the arcing stage, the temperature of the electrons exceeded that of the heavy particles significantly, and the arc deviated from LTE. Finally, the limitations of the traditional single-wavelength moiré deflection method are analyzed. The classic single-wavelength moiré deflection method, while capable of estimating heavy particle temperature in plasma, exhibits a significant error in electron density estimation compared to the dual-wavelength moiré deflection method.