Detection and Tracking of Pedestrians Using Doppler LiDAR

Abstract

Pedestrian detection and tracking is necessary for autonomous vehicles and traffic management. This paper presents a novel solution to pedestrian detection and tracking for urban scenarios based on Doppler LiDAR that records both the position and velocity of the targets. The workflow consists of two stages. In the detection stage, the input point cloud is first segmented to form clusters, frame by frame. A subsequent multiple pedestrian separation process is introduced to further segment pedestrians close to each other. While a simple speed classifier is capable of extracting most of the moving pedestrians, a supervised machine learning-based classifier is adopted to detect pedestrians with insignificant radial velocity. In the tracking stage, the pedestrian’s state is estimated by a Kalman filter, which uses the speed information to estimate the pedestrian’s dynamics. Based on the similarity between the predicted and detected states of pedestrians, a greedy algorithm is adopted to associate the trajectories with the detection results. The presented detection and tracking methods are tested on two data sets collected in San Francisco, California by a mobile Doppler LiDAR system. The results of the pedestrian detection demonstrate that the proposed two-step classifier can improve the detection performance, particularly for detecting pedestrians far from the sensor. For both data sets, the use of Doppler speed information improves the F1-score and the recall by 15% to 20%. The subsequent tracking from the Kalman filter can achieve 83.9–55.3% for the multiple object tracking accuracy (MOTA), where the contribution of the speed measurements is secondary and insignificant.

A Survey on Ground Segmentation Methods for Automotive LiDAR Sensors

“…Therefore, automotive approaches require robust and faster algorithms since the objects in the point cloud also change at higher frequencies. First approaches applied hand-crafted features and sliding windows algorithms with Support Vector Machine (SVM) classifiers for object identification, but soon were replaced by other improved methods such as 2D representations, volumetric-based, and raw point-based data, which deploy machine learning techniques in the perception system of the vehicle [ 28 ].…”

“…Due to its wide success in such domains, LiDAR sensors recently started to be adopted by the automotive industry in the perception system of the car. Their 3D point cloud can be very useful in several autonomous driving applications [ 20 , 21 , 22 ], such as obstacles, objects, and vehicles detection [ 23 , 24 , 25 , 26 ]; pedestrians recognition and tracking [ 27 , 28 ]; ground segmentation for road detection and navigation [ 29 ]; among others [ 30 ]. Because this technology works with active illumination, LiDAR sensors allow round-the-clock observations, providing accurate measurements of the vehicle’s vicinity up to hundreds of meters.…”

Learning Moving-Object Tracking with FMCW LiDAR

“…Guo et al [12] proposed a clustering and velocity estimation method based on Doppler velocity with CARLA [13] simulation. Jie Shan et al [14], [15] proposed detection and tracking methods based on hand-crafted features with Kalman filter to estimate the state of moving objects by using Blackmore FMCW LiDAR. Their methods are not based on deep learning, simple and efficient, but in low accuracy.…”

DIOR: A Hardware-Assisted Weather Denoising Solution for LiDAR Point Clouds

“…LiDAR technology is steadily improving and being applied to a wide range of applications. Accurate and precise measurements of the surroundings through a 3D point cloud representation can assist the perception systems in several tasks [3], e.g., obstacles, objects, and vehicles detection [9]- [11]; pedestrians recognition and tracking [12]; ground segmentation for road filtering [13]; among others [14]. Nonetheless, the sparse 3D point cloud of a LiDAR sensor can be subject to several noise sources, e.g., internal components, mutual interference, reflectivity issues, light, and adverse weather, which can corrupt the measurements and the output data.…”