In medical field, gaining access to blood vessels vitally important since it is the first step of many medical procedures. However, if vascular access is not established correctly, it causes complications. In this letter, we propose an algorithm that precisely segments blood vessels and automatically determines an injection point for injection robot systems using a monocular near-infrared vision system. The algorithm includes a Gaussian mixture model to determine the appropriate injection point for segmented blood vessels in a vision system and an ultrasound imaging system to estimate the depth of the blood vessel. Finally, we show the success rate of the proposed algorithm using phantom vessels.