The lighting service state of road tunnels has an important impact on traffic safety. To assess and predict traffic safety in road tunnels under different lighting service states, a traffic safety assessment system of road tunnels was innovatively established by using visual recognition clarity data from 300 field experiments, which uses the proposed traffic safety factor to assess the traffic safety of road tunnels under different lighting service states. Then, the Particle Swarm Optimization (PSO) algorithm is used to optimize the Back Propagation (BP) neural network, and a new traffic safety intelligent prediction method in road tunnels was constructed to predict the traffic safety factor under different lighting service states. The results of the study show that there is a negative correlation between simulated vehicle speed and visual clarity and a positive correlation between lighting attenuation and visual clarity. There is a negative correlation between the number of luminaire failures and the visual recognition clarity. When the lighting attenuation is below 70% or four and above luminaires fail, the lighting service states of road tunnels can seriously threaten traffic safety. The PSO-BP neural network model can accurately predict the traffic safety factor. The critical value of the traffic safety factor is 0.5965. This means that the lighting service state poses a threat to traffic safety in road tunnels when the traffic safety factor is lower than this value. The results of the study can provide the basis for developing a safe and low-carbon tunnel lighting maintenance program.