Abstract: Usually the fluorescent magnetic particle testing methods is used to detect the surface or near surface detects of ferromagnetic workpiece, but the artificial defect discrimination method is easy to cause the visual fatigue, resulting in the missing detection or misjudgment. So in this paper, from the hardware and software, the research on the fluorescent magnetic particle testing based on the machine vision was carried out. The image preprocessing method is emphatically analyzed,and a color image noise algorithm is proposed based on RGB color values. The results show that this method can effectively suppress noise interference such as magnetism, external illumination, shadow etc. It can provid the foundation and basis for subsequent research on the intelligent magnetic particle testing system.