Camera calibration and pose estimation from planes

Abstract

Camera calibration plays a key role in every computer vision application dealing with the problems of recovering a camera's geometry with respect to a 3D world reference, making 3D measurement in a captured scene or extracting 3D data from observed objects. These problems emerge in various applications such as structure from motion, robotics, augmented reality, 3D object recognition, and Simultaneous Localization And Mapping (SLAM). Using a camera as a measuring device is becoming an important trend [1], and as a consequence the need for calibrating this instrument also becomes of prime importance. Cameras are, by nature, projective devices that map our 3D world onto a 2D image. Their prime use is to create representations (i.e., images) of an environment and its actors as pictured from a particular viewpoint at a precise moment in time. Extracting 3D data from such devices is therefore inherently difficult. The camera measures the intensity of the light emitted or reflected from a certain direction; however, the depth information is lost. The projective process also implies that, in the absence of any external reference, the scale of the observed objects is undeterminable; that is an object of a given size at a given camera distance will produce the same image as an object of twice the size seen at twice the distance. In spite of these limitations, with a good understanding of the camera geometry, 3D reconstruction can become possible under specific circumstances. In this article, we review some techniques proposed in the literature to parameterize camera metric information, also referred to as camera calibration techniques. We also illustrate the use of calibrated cameras by describing two application examples involving camera pose estimation and distance estimation.