Focusing on the key research direction of the information discipline construction of our school, Associate Professor Zhang Zhenyuan, a teacher of the School of Information, systematically studied the integrated method of high-precision detection and tracking of road pedestrian targets based on commercial vehicle millimeter-wave radar sensors, aiming at the weak road target perception problem faced by intelligent driving in mountainous inclement weather scenarios. "Millimeter-Wave Radar-based Pedestrian Trajectory-Tracking for Autonomous Urban Driving" is published in IEEE Transactions, an internationally renowned journal in the field of measurement on Instrumentation and Measurement, he was recently selected as a highly cited paper by ESI.

This study takes weak pedestrian targets that are more challenging under severe weather scenarios as the research object, and carries out an integrated multi-target detection and tracking method based on sequential Monte Carlo pre-detection tracking. It adopts sequential Monte Carlo method to track the posterior probability of pedestrian target motion, and uses the combined tracking results across multiple frames to mine the relevant information of pedestrian target motion in spatio-temporal dimension. This helps millimeter-wave radar to achieve high-precision detection of pedestrian targets under low SNR environment, and then realizes mutual improvement of pedestrian weak target detection and tracking performance. In this paper, the range-velocity-azimuth spectrum of radar without threshold processing is directly used as the basis for accurate sensing of measurement data, avoiding the problem of false alarm or missing alarm caused by unreasonable detection threshold setting in traditional methods, and effectively realizing the high-precision data association between measurement and multi-target tracking trajectory under harsh environments such as rain and fog. Thus, it provides new ideas and new thinking dimensions for further improving the all-weather perception ability of automatic driving in harsh environments.

The integrated framework of "detection-tracking" for pedestrian weak targets and the experimental comparative analysis effect

The research was supported by the Youth Foundation of the National Natural Science Foundation of China (62003064), the Natural Science Foundation of Chongqing (cstc2020jcyj-msxmX0797), and the Science and Technology Project of Chongqing Education Commission (KJQN202000717). Supported by Sichuan Provincial Science and Technology Support Project (2021YJ0367) and China Postdoctoral Science Foundation (2020M683653XB). The research was completed in collaboration with Chongqing University of Posts and Telecommunications, Southwest Petroleum University, Georgia Institute of Technology (Georgia Institute of Technology) in the United States, during which Professor Huang Darong of Chongqing Jiaotong University, Professor Zhou Mu of Chongqing University of Posts and Telecommunications, Professor Ying Zhang of Georgia Institute of Technology, And Southwest Petroleum University Dr. Fang Xin careful guidance exchange and strong support.