In this study, Beijing is selected as the research area to perform wind speed correction of the aerosol optical depth (AOD) data of the 440 nm band inversion of the CE-318 solar photometer provided by AERONET (Aerosol Robotic Network) in 2014-2017. Then, the seasonal correlation analysis and modeling of the corrected daily average AOD data and the same period ground monitoring station daily average PM2.5 concentration data are conducted. Then, the visibility factor is introduced and the generalized difference method is used to construct the ternary regression model of AOD, PM2.5, and visibility in Beijing from 2015 to 2017. Finally, the data of 2014 are divided into pollution and nonpollution days for the model tests. Results show a significant linear positive correlation between AOD and PM2.5. Moreover, the seasonal differences exhibit the strongest correlation in summer, followed by that in autumn, and the weakest correlation in spring and winter. After introducing the visibility factor and eliminating the autocorrelation, the relative error of the model in the four seasons is reduced, the goodness of fit of the model significantly improved, and the relative error ranges from 21% to 31%. Compared with the previous results, the accuracy of curve fitting has been significantly improved. Moreover, the simulation effect of the model is good for low PM2.5 concentration days but poor for high PM2.5 concentration days. This study is of scientific significance for the construction of the long-term historical data of PM2.5 in Beijing.