Automatic Position Calibration Between Radar and Camera
August Bergöö och Simon Jönsson presenterar sitt examensarbete "Automatic Position Calibration Between Radar and Camera"
Abstract:A radar detector excels in poor light conditions compared to a camera, but currently there is no way to visualize the radar detections in an arbitrary camera monitoring the same scene. In this thesis, a method for unsupervised calibration of a radar-camera system is proposed. With our method, the position and orientation of the camera (in relation to the radar) can be calculated automatically. The calibration also provides optical (intrinsic) parameters of the camera, such as the focal length and the radial distortion coefficient. The calibration process consists of collecting data over time by detecting moving objects. Then, point correspondences between the radar and camera are established on which a homography is calculated by a combination of Random Sample Consensus (RANSAC) and Sequential Least Squares Programming (SLSQP). After the calibration, the radar detection can be overlaid in the correct place in the camera view, giving visual guidance to camera operators.