Tracking and data association
An Experimental Comparison of Range Image Segmentation Algorithms
IEEE Transactions on Pattern Analysis and Machine Intelligence
Mean Shift: A Robust Approach Toward Feature Space Analysis
IEEE Transactions on Pattern Analysis and Machine Intelligence
Tracking a Variable Number of Human Groups in Video Using Probability Hypothesis Density
ICPR '06 Proceedings of the 18th International Conference on Pattern Recognition - Volume 03
Statistical Multisource-Multitarget Information Fusion
Statistical Multisource-Multitarget Information Fusion
Convergence results for the particle PHD filter
IEEE Transactions on Signal Processing
The Gaussian Mixture Probability Hypothesis Density Filter
IEEE Transactions on Signal Processing
3D object tracking with a high-resolution GPU based real-time stereo
Proceedings of the 27th Conference on Image and Vision Computing New Zealand
| Hi-index | 0.00 |
This paper presents a random finite set theoretic formulation for multi-object tracking as perceived by a 3D-LIDAR in a dynamic environment. It is mainly concerned with the joint detection and estimation of the unknown and time varying number of objects present in the environment and the dynamic state of these objects, given a set of measurements. This problem is particularly challenging in cluttered dynamic environments such as in urban settings or marine environments, because, given a measurement set, there is absolutely no knowledge of which object generated which measurement, and the detected measurements are indistinguishable from false alarms. The proposed approach to multi-object tracking is based on the rigorous theory of finite set statistics (FISST). The optimal Bayesian multi-object tracking is not yet practical due to its computational complexity. However, a practical alternative to the optimal filter is the probability hypothesis density (PHD) filter, that propagates the first order statistical moment of the full multi-object posterior distribution. In contrast to classical approaches, this random finite set framework does not require any explicit data associations. In this paper, a Gaussian mixture approximation of the PHD filter is applied to track variable number of objects from 3D-LIDAR measurements by estimating both the number of objects and their respective locations in each scan. Experimental results obtained in marine environments demonstrate the efficacy and tracking performance of the proposed approach.