Régis Lherbier

Feature Extraction in Scanning Laser Range Data Using Invariant Parameters: Application to Vehicle Detection

This paper presents a feature extraction method in scanning laser range data. Many authors have studied this problem by proposing solutions that rely on a modeling of the scene in Cartesian coordinates. These methods are based on the computation of the interscan distance between two consecutive measurements, which, in practice, is not very easy to estimate. Our proposed method, i.e., segmentation using invariant parameters (SIP), deals with laser measurements in natural coordinates, which avoids any preprocessing stage that could modify the measurement noise statistics. This approach is founded on the use of an invariant description of the feature and leads to the definition of a criterion of line-segment detection that only depends on the sensor intrinsic parameters.

Automatic feature extraction in laser rangefinder data using geometric invariance

This paper presents a feature extraction method in scanning laser rangefinder data. Whereas many popular methods use Cartesian coordinates to detect features, the proposed method use a geometric invariant in natural (polar) coordinates. It leads to a membership condition that only depends on the sensor properties (angular resolution and range measurement error).

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

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.

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

This paper presents a method for joint detection and tracking of vehicles with a scanning laser rangefinder. The lidar measurements of an object have the particularity to be spatially distributed, which generally leads to a detection step before any tracking. Differently, the proposed method relies on the raw measurement processing without any detection step, which improves the overall performance in multiobject tracking while providing good estimation accuracies. The solution uses the sequential Monte Carlo methods by incorporating the geometric invariant of the objects of interest (vehicles). This approach also offers an efficient solution to the problem of multitarget tracking by integrating naturally the track management in the filtering process.

A PHD approach for multiple vehicle tracking based on a polar detection method in laser range data

This paper presents a detection and tracking approach of multiple vehicles in scanning laser range data. The proposed solution relies on a new detection method based on object geometric invariant that uses the raw measurements directly in polar coordinates. The multitarget management problem is solved in the PHD framework by a particle filter.

Use of contextual information by Bayesian Networks for multi-object tracking in scanning laser range data

This paper presents a new method to improve the perception of the environment of a vehicle. The aim here is to detect the vehicles using a scanning laser range finder and track the detected objects at each time. This contribution can be considered as an element of a global vehicle-to-vehicle (v2v) surveillance system where the on-board system receives warnings from the other local systems. It allows to extend the effective surveillance field and as a consequence to provide a faster reaction of the vehicle (collision avoidance or mitigation). To deal with this multi-target tracking problem, we focus on the Joint Probabilistic Data Association (JPDA) methods. Their particularity lies in their ability to take into account the probabilistic characteristics of the detector (detection and false alarm probabilities). Whereas in many works the detection probability is set up once, our contribution is to propose a method that dynamically estimates for each object the detection probability using the contextual information modeled by a Bayesian Network (BN-JPDAF).

A multi-sensor validation approach for human activity monitoring

In this paper, we present a sensor validation approach for human activity monitoring. The sensor validation indicator is used for controlling the tracking process by a reactive strategy. The validation is represented by a confidence associated to each current tracked object and for each sensor and integrates static and dynamic contextual information. This confidence value is estimated by a Bayesian Network, which combines a set of heterogeneous information. The validation indicator takes into account the notion of geometrical visibility and the size of the detected object. We have tested this sensor validation approach in an indoor environment by using two complementary sensors for human activity monitoring. The first sensor is an industrial stereo camera (Triclops) and the second is a laser scanner (Sick LMS).

Multiple target detection and tracking by interacting joint probabilistic data association filter and bayesian networks: Application to real data

This paper proposes an algorithm of multiple target detection and tracking on road, developed for the laserscanner data. It is based on Bayesian networks for calculating the detection probability of target used in a JPDA filter. We propose a method based on the integration of detection probability of target in the JPDA filter, in which the joint probabilities of associations are calculated for multiple target. It then solves the problem of founding one or more observation in more than one gate of target by constructing hypothesis. A Bayesian network is used to determine this probability of detection. It includes a target model and takes into account the contextual information such as the size of the target, the distance between targets and the center of the sensor characteristics (position, measurement uncertainty). The method is applied to real data provided by a scanning laser sensor with multiple targets. as well as the possibility that an object enters and leaves the scene.

Bayesian Networks and Probabilistic Data Association Methods for Multi-Object Tracking: Application to Road Safety

This paper presents a Bayesian network-based approach to multisensor multitarget detection and tracking problem. The aim here is to propose an improvement of the probabilistic data association approach that takes into account contextual information about the scene. This information is modeled by a Bayesian network that allows a dynamic estimation of the detection probability of the PDA. Our approach is then applied to synthetic data from scanning radar that is mounted on a moving vehicle. The aim is to detect the surrounding objects and track them through the sequence.

A generic particle filtering approach for multiple polyhedral object tracking in a distributed active sensor network

This paper presents a method for multi-target tracking in a multisensor system composed of several distributed active sensors (rangefinders). The final goal is to deliver a complete reconstruction of the environment of a vehicle. The originality of this work relies on the joint exploitation of geometric invariance properties of objects avoiding any loss of optimality and on an efficient management of transitions when the objects of interest move from one field-of-view to another.