For a fringe projection profilometry, the projector calibration is the key step to ensure the measurement accuracy of the system. However, conventional methods for the projector calibration are usually involved a complex system model and a calibration process. Therefore, an approach for accurate projector calibration based on the Grey Wolf Optimization (GWO) algorithm and the multi-layer perceptron (MLP) is proposed to simplify the process and improve the generalization. In this method, in order to compensate the errors of projected image coordinates of the calibrated corners obtained by using the projective transformation matrix, the parameters of the MLP network are optimized by using the GWO to construct the GWO-MLP network. Then, it is trained to compensate the errors of the coordinates of characteristic corners and is available to calibrate the projector. The experiments verify the feasibility and effectiveness of the proposed method.