The particle streak velocimetry (PSV) method based on binocular vision and single-frame multiple exposure was proposed in this paper. To be specific, the velocity direction was judged by the combination of long and short exposure and the gray-level fitting was used to identify the trajectory features. In addition, the trajectory midpoint was employed to calculate the particle velocity in order to improve the measurement accuracy. This method adopted nearest neighbor matching and angle of inclination constraints to match and sort multiple trajectories of the same particle, epipolar constraints to accomplish binocular matching based on trajectory feature points, and B-spline fitting of the three-dimensional reconstructed feature points to calculate the three-dimensional velocity of particles. The multi-exposure PSV method can simplify binocular matching. As a result, an experimental system and the corresponding image processing flow were formed for binocular multi-exposure PSV. We applied the small luminous holes to verification experiments at different three-dimensional velocities and analyzed the influence of different trajectory feature points on the velocity measurement. The results show that the average relative error obtained by the trajectory midpoint is less than 4.2%, and the standard deviation is small, demonstrating the advantage of velocity calculation with the midpoint rather than with the end. The actual measurement of the flow field in a water jet test bench indicates that this method can be reliably used for three-dimensional flow field measurement.