A Simplified Active Calibration algorithm for Focal Length Estimation
This provides a simplified calibration method for robotic and dynamic surveillance environments where cameras move and zoom freely, though it appears incremental as it builds on existing active calibration approaches.
The paper tackles camera focal length estimation by introducing linear mathematical formulations that require only one point correspondence between images taken before and after a degenerate camera rotation, enabling calibration during free movement and zooming. The method achieves low error rates in evaluations on simulated and real-world images.
We introduce new linear mathematical formulations to calculate the focal length of a camera in an active platform. Through mathematical derivations, we show that the focal lengths in each direction can be estimated using only one point correspondence that relates images taken before and after a degenerate rotation of the camera. The new formulations will be beneficial in robotic and dynamic surveillance environments when the camera needs to be calibrated while it freely moves and zooms. By establishing a correspondence between only two images taken after slightly panning and tilting the camera and a reference image, our proposed Simplified Calibration Method is able to calculate the focal length of the camera. We extensively evaluate the derived formulations on a simulated camera, 3D scenes and real-world images. Our error analysis over simulated and real images indicates that the proposed Simplified Active Calibration formulation estimates the parameters of a camera with low error rates.