A particle filtering approach for joint vehicular detection and tracking in lidar data

Abstract

This paper presents a method for joint detection and tracking of vehicles in scanning laser range data. Many methods use a solution that processes the raw data in a detection procedure and then tracks the detected object in an association/tracking procedure. The proposed approach uses a preclustering stage (SIP) as an input of the tracking process that allows to manage the displacement of the center-of-gravity and the changes in the apparent shape from object and motion modeling. The global problem is then described using a state-space modeling which is solved by a nonlinear filtering method.

Obstacle detection and tracking method for autonomous vehicle based on three-dimensional LiDAR

“…The selection of tracking point has a great impact on the tracking result of moving obstacle. When using HDL-64E LiDAR to track the moving obstacle, point-cloud data combined with the prior shape knowledge of the moving obstacle can deduce the moving obstacle's center position, 20,21 but in fact, the shapes of all the moving obstacles do not strictly meet the specific models, and because of the obstacle's occlusion and self-occlusion, the deduced center position cannot represent the vehicle's position precisely. We improve Xiao et al's method, 22 and when the moving obstacle is in different regions, different optimum tracking point is selected, and when the moving obstacle switches from a region to another region, the length and width of the moving obstacle are taken into account.…”

“…The optimal estimationX kþ1 at time k þ 1 can be obtained through equation (21), where K kþ1 is the Kalman gain which can be obtained through equation ( 22) at time k þ 1, and Z kþ1 is the actual observation of moving obstacle at time k þ 1.…”

A Model-Based Joint Detection and Tracking Approach for Multi-Vehicle Tracking With Lidar Sensor

“…First detection results have been presented in [10] and the gain is shown regarding conventional detection approaches (Split-and-Merge, Incremental, Ransac), which are generally based on a Cartesian modeling. Here, we propose a modelbased approach on this framework which avoids the detection step [22]. As a consequence, the modeling of the filtering problem must be rewritten by using the aggregated data.…”

“…In this case, the SIP detector [22] is used to initialize the filter according to one of the prior defined object classes (trucks, large size car, small size car).…”

“…In this kind of method, a major difficulty so lies in the needing of a compatibility between the raw measurements aggregation and the tracking process. We proposed in [21] a detection method of line segment (SIP: Segmentation using Invariant Parameters) based on invariant parameters of the shape and first results in a single target case have been presented in [22]. It showed the interest of dealing directly with the raw measurements in terms of robustness towards outliers and minimization of the nondetection probability.…”