An objective data processing method is presented for the calibration of sun photometers CE318 manufactured by the CIMEL Company (France) by using the Langley method and conventional observation data. There are four steps: (1) the observational data from sun photometers are processed with the Langley method in order to acquire calibration coefficients, the linear fitting correlation coefficients for the Langley method, the aerosol optical depth at 550 nm and so on; (2) calibration coefficients are screened through three thresholds: correlation coefficients at 440 nm （≥0.99）, aerosol optical depth at 550 nm （≤0.15） and the number of data sets for linear fitting in 12 hours （≥8）; (3) calibration coefficients are screened again according to voltage signal images and the maximum at noon; (4) the annual mean calibration coefficients are obtained from the remain data. The data from four sun photometers at Urumqi, Tazhong, Hotan and Hami in Xinjiang from 2002 to 2005 are processed according to the above four procedures. The final mean calibration coefficients are compared with the calibration results acquired through intercomparison with the reference sun photometer in Beijing in 2005 and standard coefficients provided by the CIMEL Company. Biases in these data at three sites are all less than 8.5% (except for the error of Hotan Station), which shows that the procedures for calibration are feasible.