A method for fast and non-destructively detecting soil pH by using the color features in multispectral images is proposed. First, a 2 CCD multispectral imager is used to obtain the R, G, B and NIR images of each soil sample respectively. The multispectral images are converted in color space, i.e. the RGB color spaces are converted to the HSV color space and the Lab color space. Then, the color features of the multispectral images are extracted in different color spaces. Finally, the extracted color features are used as input variables, so as to establish a prediction model of soil pH by using PLS and LS-SVM algorithms. The experimental result shows that it is feasible to detect soil pH by using the multispectral image technology. The best result is given by the prediction model established. Its determination coefficient (R2) and root mean square error (RMSEP) value are 0.91 and 0.34 respectively.