Multi‐view convolutional neural network‐based target classification in high‐resolution automotive radar sensor

In this study, a target classification method based on point cloud data in high-resolution radar sensor is proposed. By using multiple antenna elements arranged horizontal and vertical directions, pedestrians, cyclists vehicles can be expressed as the three-dimensional (3D) space. To perform spatial characteristics (i.e. length, height width) of target, 3D orthogonally projected onto xy, yz zx planes, respectively, three types images are generated. Then, multi-view convolutional neural network (CNN)-based classifier those inputs designed. end, for synthesising detection results directions series or parallel The proposed learn by viewpoints. Compared to CNN-based that uses only result single plane input, shows 4.5%p higher accuracy terms with lowest accuracy. addition, CNN structure improved performance shorter training time compared well-known deep learning methods image classification.