A Point Cloud-Based Multi-class Car Classification by Using mmWave Radar and Deep Learning Models

Object detection and classification have shown a great interest in the recent past. It helps to enhance efficiency, security, and safety in broad applications, including traffic management, surveillance, autonomous driving, and pedestrian safety. Camera-based detection is used, but it is not promising during adverse weather conditions. In order to address this problem, millimeter wave (mmWave) radar-based multi-class car detection and classification are used in this paper. The radar used here in hardware implementation is manufactured by Texas Instruments IWR1843BOOST. This system used 3D point cloud data from mmWave radar. Objects are categorized into three groups of automobiles: SUV, sedan, and hatchback. Furthermore, we employed Deep Neural Network (DNN), Convolutional Neural Network (CNN), and Residual Network (ResNet) models for classification purposes. After training both models in the dataset, ResNet achieved the highest accuracy of 80%.