used 3D point clouds generated by a planar phased array FMCW radar to detect different human motions. Based on a new feature vector, it used a multivariate Gaussian mixture model (GMM) for radar point cloud segmentation in an unsupervised learning environment, i.e., “point-by-point” classification. used a high-resolution MMW radar sensor to obtain a radar point-cloud representation for traffic surveillance scenes. In a related task based on imaging radar, Feng et al. In order to estimate the direction of arrival, examples were executed by combining them. used a new antenna array device that provides the ability to measure angles in azimuth and elevation. Lastly, road edge height estimation, drain detection and parking lot detection were accomplished using this radar. The multiple-input multiple-output (MIMO) technique and binary phase shift keying (BPSK) coding were used for transmitting signals to obtain elevation information. proposed a novel 4D radar that operates at 79 GHz with 1.6 GHz bandwidth and uses frequency-modulated continuous wave (FMCW). In terms of imaging radar hardware, Li et al. However, research on related algorithms is still in the initial stage. The imaging radar can produce LIDAR-like point-cloud data, contain rich Doppler information and it has all-weather characteristics. With the advent of a new generation of 4D high-resolution imaging radars, promising applications have been seen. Furthermore, it lacks object height information and serves only as a last line of defense in most autonomous driving systems, acting as an advanced warning. The traditional MMW radar for commercial vehicles is affected by its resolution, making it challenging to perform object classification tasks. It is indisputable that cameras and LIDAR fail to varying degrees in the rain, snow and fog and under operating conditions such as bright light and darkness, while the MMW radar is indispensable as it shows strong robustness under bad conditions. The mainstream sensors in the environmental sensing module mainly consist of cameras, LIDAR and MMW radar. Among them, environment perception is significant and its good performance directly affects the downstream modules. Autonomous vehicles mainly consist of several modules such as environment perception, path planning and decision control. In recent years, autonomous driving technology has developed rapidly and received wide attention. The experimental results show that our proposed method achieved an overall classification accuracy of 94.9%, which is more suitable for processing radar point clouds than the popular deep learning frameworks and shows promising performance. We generated an imaging radar classification dataset and completed manual annotation. The algorithm takes the attention mechanism as the core and adopts the combination of vector attention and scalar attention to make full use of the spatial information, Doppler information, and reflection intensity information of the radar point cloud to realize the deep fusion of local attention features and global attention features. In this paper, we propose an object classification network named Radar Transformer. It has high azimuth and elevation resolution and contains Doppler information to produce a high-quality point cloud. Thus, the concept of a new generation of four-dimensional (4D) imaging radar was proposed. Traditional commercial automotive radars are limited by their resolution, which makes the object classification task difficult. Automotive millimeter-wave (MMW) radar is essential in autonomous vehicles due to its robustness in all weather conditions.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |