Neural network training method and global calibration method for FOD radar and camera

The invention provides an FOD radar and camera neural network training method and a global calibration method. The method comprises the following steps: detecting the distance and azimuth angle of each calibration point relative to the FOD radar through the FOD radar; acquiring installation data of the FOD radar and the camera; and through the trained neural network model, obtaining the predicted position of the target in the image picture and matching the predicted position with the actual coordinate position detected by the camera, or obtaining the predicted position of the target in the FOD radar coordinate system and matching the predicted position with the actual coordinate position detected by the FOD radar. According to the method, a neural network technology is adopted, calibration is carried out at multiple orientations and multiple distance sections at the same time, a set of calibration parameters suitable for the whole detection area is obtained, errors caused by the fact that the ground is not ho