More reliable classification of radar data from dynamic scenarios

摘要

Procedure (100) for classifying radar data (2) obtained by recording radar radiation emitted from a transmitter and reflected by at least one object (20) with at least one detector, using the steps:Provide (110) radar data (2) containing observations recorded at different times (2a-2c) of a scene (1);Determine (120) at least one part (2') of the radar data (2), which is rotated and/or scaled in at least one of the observations (2a-2c) compared to at least one other of the observations (2a-2c);Determine (130) a fixed point (3) of rotation and/or scaling;Transform (140) at least a two-dimensional representation (4a-4c) of at least one part of the observations (2a-2c) into logarithmic polar coordinates with the determined fixed point (3) as origin;Images (150) of at least one transformed two-dimensional representation (4a'-4c') on at least one class (6a-6d) of a predetermined classification (6) by at least one classification (5), comprising a neural network with at least one folding layer.Procedure (200) for training a classifier (5).