In traffic monitoring, weather conditions can disrupt the accuracy of detection systems that use LiDAR and cameras. However, radar remains robust in adverse weather scenarios. In our project, we proposed a systematic method for fusing LiDAR and radar sensors. First, we developed a demo integrated sensor system using the TI-1843 radar and Vanjee L-128 LiDAR. Second, we used this system to collect traffic data in two different cities under both sunny and rainy conditions. Finally, we demonstrated that fusing the collected data enhances detection accuracy.
Fig.1 Data collecting with our proposed fusion sensor.
Fig.2 Traffic data collected by our fusion multi-modality sensor. The left images shows the data that collected with a moving vehicle. The right image illustrate the data collected from road side view. In the left picture, the red and green points indicate the radar points and LiDAR points. To the right of them are radar signals images with blue and yellow color which shows the intensity of the signal. There is also a RGB iamge to show the environment of that real time. In the left iamge, the four sub-figures is RGB image (left up), LiDAR points (right up), radar range-doppler signal (left down) and overlap display of lidar points(white) and radar range-azimuth heatmap (right down). More details can not be opened as the privacy of this project.
Fig.3 Radar cube data formation visualization (left). Radar data processing flow (right).
According to the standard radar processing pipeline, the radar data is represented as a complex 3D data cube. Its three axes are defined as fast time (range), slow time (Doppler), and antenna (angle), as illustrated in Fig. 3. By following this processing procedure, one can locate detected objects and measure their speed.