In this paper, a vision-based measuring device is proposed and experimentally demonstrated to be an accurate, flexible, and low-cost tool for the kinematic calibration of parallel mechanisms. The accuracy and ease of use of the proposed vision sensor are outlined, with the suppression of the need for an accurate calibration target, and adequacy to the kinematic calibration process is investigated. In particular, identifiability conditions with the use of such an exteroceptive sensor are derived, considering the calibration with inverse or implicit models. Extensive results are given, with the evaluation of the measuring device and the calibration of an H4 robot. Using the full-pose measurement, an experimental analysis of the optimal calibration model is achieved, with study of the kinematic behavior of the mechanism. The efficiency of the provided method is thus evaluated, and the applicability of vision-based measuring devices to the context of kinematic calibration of parallel mechanisms is discussed.