In the biological micromanipulation experiment, it is of great significance to capture biological targets safely and stably. The traditional manual operation has problems of low efficiency and poor repeatability. To address these issues, a set of biological automatic capture system for micromanipulation is proposed. Through visual feedback and closed-loop control, the micropipette is used to automatically aspirate biological objects in a liquid growth medium. Firstly, the particle swarm optimization algorithm is used to optimize the traditional image segmentation algorithm to realize the synchronous real-time tracking of the biological target in the field of view and the position of the pipette. Then, the dynamic model of micropipette capture process is formulated, the nonlinear disturbance observer is used to suppress the uncertainty of model parameters and environmental disturbance, and the closed-loop control system is established. Finally, the performance of the system is evaluated by experiments. Results show that the average image segmentation time of the system is 81. 05 ms, the average capture time is 1. 85 s, the average maximum error of capture is 0. 34 mm, and the capture success rate is 94% . The experiments show that the system can accurately, quickly and undamaged capture biological objects in the environment with trace interference in different light sources and visual fields, which has good robustness. The potential applications of this method include vitrification of embryos, embryonic stem cell transplantation, blastomere biopsy, cell mechanical property detection, etc.