There is no difference in the evaluation of the direct-view low-light-level night vision system for the three-generation and super-second generation image intensifiers. The simulation results are not ideal under the ultra-low illumination of 10^-4 lx. Therefore, three optimizations were made for the model: the transfer function of the human eye was added into the transfer function model of the system; the average photoelectron number and photoelectron contrast formula generated by the scene at the photocathode were corrected based on the spectral matching between the scene and the photocathode; the background noise term was added to the noise source. The improved apparent distance model was verified by the experimental results of the helmet system in field conditions. The results show that the improved model can effectively distinguish the difference between the recognition distances of the super-second generation and three-generation night vision systems, the calculation results are closer to the actual measurement results.