Structured-light-based 3D multi-directional sensors are widely used due to their flexible field of view (FOV), which is a significant benefit. Nevertheless, existing 3D multi-directional sensors based on linear lasers are only capable of measuring a slice of a 3D scene at low resolution. In addition, other 3D multi-directional sensors are capable of measuring 3D scenes by shooting 2D structured light patterns in different directions through curved mirrors. However, the projected and captured structured light patterns are blurred due to the optical aberration property of curved mirrors, resulting in inaccurate 3D reconstruction. This paper proposes a 3D, high-resolution, high-accuracy, real-time multi-directional sensor that utilizes pyramid mirrors to decompose the FOVs into different directions and reduce the optical aberration phenomenon. The mathematical model of the 3D sensor is presented, and the FOV is analyzed. A dual-frequency phase-shifting fringe pattern is used for 3D reconstruction in real time. Moreover, the sensor is calibrated using a planar circle calibration gauge and the Householder reflection constraint. Finally, the experimental results demonstrate that the measurement error of the 3D multi-directional sensor is less than 1mm and 0.03rad, thus verifying the method's feasibility.