Using the AKARI mid-infrared all-sky survey catalogue, we are searching for debris disks which are important objects as an observational clue to on-going planetary system formation. Debris disk candidates are selected through a significant excess of the measured flux over the predicted flux for the stellar photospheric emission at $18{mu}m$. The fluxes were originally estimated based on the near-infrared spectral energy distributions (SEDs) of central stars constructed from the 2MASS J-, H-, and Ks-band fluxes. However, we found that in many cases the 2MASS photometry has large errors due to saturation in the central part of a star image. Therefore we performed follow-up observations with the IRSF 1.4m near-infrared telescope in South Africa to obtain accurate fluxes in the J-, H-, and Ks-bands. As a result, we have succeeded in improving the SEDs of the central stars. This improvement of the SEDs allows us to make more reliable selection of the candidates.