A dynamic floc measurement system was developed to analyze flocs during sedimentation and automatically extract floc characteristics without damaging them. An optimized algorithm was proposed to remove fuzzy flocs, and visual inspection verified that this algorithm achieved satisfactory separation performance between well-defined and fuzzy flocs. This program was used to study the development of coal flocs and showed that the flocs grew rapidly at the beginning of flocculation. The floc size at equilibrium at low stirring velocities was larger than those at high stirring velocities. A high stirring velocity can cause irreversible damage to the flocs, and those flocs cannot be restored to their original size by decreasing the stirring velocity. In addition, the fractal dimension decreased with increasing time, indicating that there is an optimal stirring velocity which can be selected to enhance the size and structure of flocs. This research provided an effective way to study the characteristics of flocs, and the change process of the flocs under different stirring velocities was investigated in detail to understand the growth, breakage, and regrowth process.