Cell rotation has great significance for cell manipulation, which is applied to intracytoplasmic sperm injection, preimplantation genetic screening and diagnosis, somatic cell nuclear transfer, etc. In this letter, an automatic cell rotation method is proposed based on real-time detection and tracking of the polar body. Image segmentation and visual tracking are combined to detect the polar body accurately and robustly during cell rotation. Both out-of-plane and in-plane rotation are realized by utilizing a pair of standard glass micropipettes. In experiments, two kinds of oocytes, which are quite different in appearance, are used to validate the system. For mouse oocytes, the robotic system achieved a 98% success rate of out-of-plane rotation and a 93% success rate of in-plane rotation. For porcine oocytes, the robotic system achieved an 80% success rate of out-of-plane rotation and a 93% success rate of in-plane rotation. The average time of cell rotation is approximately 9.1s for mouse oocytes and 12.8s for porcine oocytes. The experiment results demonstrate that the proposed method is effective, stable and high speed for automatic cell rotation in 3D.