ICSI (Intra-Cytoplasmic Sperm Injection) has attracted research interest from both biological and engineering groups. This technique is used to create genetically identical offsprings of many different species (mice, rats, cattle et al.). We focus on a relatively recent technology for ICSI, called Ros-Drill© (Rotationally Oscillating Drill). In this paper, we present a procedure to automate a critical part of the ICSI operation: initiation of the rotational oscillation phase. (The visual feedback is on the real-time visual monitoring of the tip of the injection pipette. First, the real-time feature extraction in visual feedback is explained). The proposed algorithm which does this is split up into two parts: off-line and on-line. In the off-line part a proper thresholding grey level is determined. In the on-line part a realtime algorithm tracks the tip of the injecting pipette. A predetermined, species-specific penetration depth is used to initiate the rotational oscillation action of the drill. This real-time visual feedback control is compared with an earlier curvature-based analysis. Algorithmic stages as well as the outcomes from the concept validation tests are presented.