A triple-frequency color-encoded fringe projection profilometry based on empirical mode decomposition is presented to measure the dynamic objects in real time. Sinusoidal fringe patterns of three frames with different frequencies are encoded in red, green, blue (RGB) channels, respectively, projected on the objects by projector at the same time, and the deformed fringe patterns are captured by CCD at another angle. The background components are eliminated after subtracting the red channel′s component from both blue and green components. Then empirical mode decomposition is used for color decoupling and separating the highest, medium and lowest fundamental frequency components. Phases are demodulated by using Fourier transform, and the warpped phase demodulated from the highest fundamental frequency components is unwrapped by three-step phase unwrapping with variable precision algorithm. The simulation experiment, in which the standard deviation of phase demodulation is less than 0.0417 rad, shows the method has high precision. Furthermore, the method is verified to be effective by contrast experiments and experiment on facial expression. It implements phase demodulation and phase unwrapping with high precision based on single snapshot, thus it is very effective to measure 3D surface contour of dynamic objects with high precision.