While several automatic manipulation techniques have been developed for optical tweezers system recently, the measurement of the velocity of cell is required and the interaction between the cell and the manipulator of laser source is usually ignored in these formulations. Although the position of cell can be measured by using a camera, the velocity of cell is not measurable and usually estimated by differentiating the position of cell, which amplifies noises and may induce chattering of the system. In addition, it is also assumed in existing methods that the image Jacobian matrix from Cartesian space to image space of the camera is exactly known. In the presence of estimation errors or variations of depth information between the camera and the cell, it is not sure whether the stability of the system could still be ensured. In this paper, vision based observer techniques are proposed for optical manipulation to estimate the velocity of cell. Using the proposed observer techniques, tracking control strategies are developed to manipulate biological cells with different Reynolds’ number, which do not require camera calibration and measurement of the velocity of cell. The control methods are based on the dynamic formulation where the laser source is controlled by closed-loop robotic manipulation technique. The stability is analysed using Lyapunov-like analysis. Simulation and experimental results are presented to illustrate the performance of the proposed cell manipulation methods.