This paper proposes a close-range relative navigation method based on computer vision where an autonomous underwater vehicle (AUV) estimates its relative position to a marker made lights arranged in a determined geometric disposition which can be attached to another AUV for trajectory tracking tasks. The vision algorithm segments sources of light to detect candidates which can be a part of the marker used by this method, those light candidates are filtered so the lights that belong to the marker can be recognized among other sources of light or reflections on the surface. The estimation is done with a calibrated monocular camera, using the known distance between each pair of lights and the field of view of the camera. In order to confirm the effectiveness of the developed vision algorithm, a set of experimental results are conducted in a BlueRov 2 underwater vehicle, where in a control ground station the vision algorithm is implemented as well as a a Super-Twisting sliding mode controller.