For the rapid detection of the total organic carbon (TOC) content and cation exchange capacity (CEC) in soil, visible/near infrared spectra (Vis/NIR) of 300 soil samples were analyzed. The algorithm of fast independent component analysis (FastICA) was used to decompose the data of Vis/NIR spectrum, and their independent components and the mixing matrix were obtained. Then, the calibration model with three-level artificial neural networks structure was built by using Back-Propagation (BP) algorithm. Genetic algorithm was used to revise the weights of neural networks to quicken the rate of convergence and overcome the problem of falling easily into local minimums, and finally the ICA-GA-BP model was built. The models were used to estimate the content of TOC and CEC in soil samples both in calibration set and predicted set. Correlation coefficient (R2) of prediction and root mean square error of prediction (RMSEP) were used as the evaluation indexes. The results indicate that the R for the prediction of TOC content and CEC can both reach 0.98. These indicated that the results of analysis were satisfiable based on ICA method, and offer a new approach to the fast prediction of components’ contents in soil.