This study explores and preliminarily validates an online calibration method for Augmented Reality Optical See-Through Head-Mounted Displays. One of the major obstacles with those devices concerns the complexity and the unreliability of the calibration procedures for a correct virtual-to-real alignment, a key aspect of medical and industrial settings. Specific simplified procedures calibrate Commercial Optical See-Through devices, but these are mainly based on the adjustment of the single degree of freedom associated with the user's interpupillary distance. The presented work aims to provide a calibration method that can improve the performance and the accuracy of the user with precise control of the rendering cameras intrinsic and extrinsic parameters on the top of the device's default calibration system. The study was carried out using Microsoft HoloLens 2, which is one of the most advanced Optical See-Through Head-Mounted Displays, and we asked users to complete a targeting task within their peripersonal space after calibrating the device. Preliminary results show that the proposed method improves the alignment task by approximately 50 % in terms of mean error and by approximately 35 % in terms of maximum error.