The concentration of carbon dioxide in the atmosphere has been increasing continuously. This is a big concern with respect to the environment and climate changes and so on. A kind of ground-based low-resolution remote sensing system and a real-time spectral inversion method are presented for the real-time remote sensing column concentration of carbon dioxide. Based on continuous observation of this system in Hefei, the total atmospheric transmittance is real-timely obtained from the solar absorption spectrum. The column concentrations of carbon dioxide and oxygen are inversed from total atmospheric transmittance by line-by-line nonlinear least squares spectral inversion algorithm. The column concentration of oxygen is used as internal standard function to obtain the column average volume mixing ratio of carbon dioxide in total dry air column, and the precision is better than 3%.The average volume mixing ratio of carbon dioxide in dry-air column, from 12: 00 to 15:00 on September 25, 2012, is compared with observation results by Japanese greenhouse-gases satellite over the site in this period, showing that the relative deviation is less than 1%. Obviously, it is an effective way to measure greenhouse gases concentration with high precision and accuracy.