Autocalibration of monocular cameras for autonomous driving scenarios

Object identification, detection and distance calculation are crucial topics for the autonomous driving world. Different technologies are used in the automotive field to reach the expected results. LIDAR and radar technologies are widely used for distance measurement, but they are quite expensive and rather not able to cover all the use cases (e.g. road lanes or traffic signs identification). For this reason, those sensors are often used in conjunction with algorithms where also camera data is integrated. A possible alternative to LIDAR and radars is the usage of stereo cameras. The stereo camera is a sensing technology using two cameras to capture images. Since stereo cameras acquire images with multiple cameras, they estimate the distance between the camera itself and the object surface by exploiting points triangulation, which is not possible with a monocular camera system. In this way, stereo vision has a broad range of applications. This work aims at proposing a method to self-calibrate four monocular cameras in order to retrieve with an acceptable approximation the positions of the cameras in the real-world coordinates. This will be an important step in order to exploit future achievements with the growing availability of stereo cameras.