Abstract:Defining a multi-target motion model, which is an important step of tracking algorithms, can be very challenging. Using fixed models (as in several generative Bayesian algorithms, such as Kalman filters) can fail to accurately predict sophisticated target motions. On the other hand, sequential learning of the motion model (for example, using recurrent neural networks) can be computationally complex and difficult due to the variable unknown number of targets. In this paper, we propose a multi-target filtering and tracking algorithm which learns the motion model, simultaneously for all targets, from an implicitly represented state map and performs spatio-temporal data prediction. To this end, the multi-target state is modelled over a continuous hypothetical target space, using random finite sets and Gaussian mixture probability hypothesis density formulations. The prediction step is recursively performed using a deep convolutional recurrent neural network with a long short-term memory architecture, which is trained as a regression block, on the fly, over "probability density difference" maps. Our approach is evaluated over widely used pedestrian tracking benchmarks, remarkably outperforming state-of-the-art multi-target filtering algorithms, while giving competitive results when compared with other tracking approaches: The proposed approach generates an average 40.40 and 62.29 optimal sub-pattern assignment (OSPA) errors on MOT15 and MOT16/17 datasets, respectively, while producing 62.0%, 70.0% and 66.9% multi-object tracking accuracy (MOTA) on MOT16/17, PNNL Parking Lot and PETS09 pedestrian tracking datasets, respectively, when publicly available detectors are used.

Kernel-based Learning of Birth Process from Evolving Spatiotemporal RFS Data Stream in SMCCPHD Filter for Multi-Target Tracking

Estimations of Time-Varying Birth Cardinality Distribution and Birth Intensity in Gaussian Mixture CPHD Filter for Multi-Target Tracking

An Efficient Data-Driven Particle PHD Filter for Multitarget Tracking.

Improved SMC-PHD Filter for Multi-Target Track-Before-Detect.

Current Statistical Model Probability Hypothesis Density Filter For Multiple Maneuvering Targets Tracking

Novel multiple-model probability hypothesis density filter for multiple maneuvering targets tracking

Large Margin Object Tracking with Circulant Feature Maps

A novel SMC-PHD filter for multi-target tracking without clustering

An Efficient Measurement-Driven Sequential Monte Carlo Multi-Bernoulli Filter for Multi-Target Filtering

A Robust SMC-PHD Filter for Multi-Target Tracking with Unknown Heavy-Tailed Measurement Noise

Adaptive Multi-Hypothesis Marginal Bayes Filter for Tracking Multiple Targets

Measurements Optimal Assigned GM-PHD Multi-target Tracking Algorithm

A Multilevel Homotopy MCMC Sequential Monte Carlo Filter for Multi-Target Tracking

An Adaptive Track Initiation Method Adopting Multiscan Spatiotemporal Information

A Novel Video Object Tracking Approach Based on Kernel Density Estimation and Markov Random Field

Cluster target tracking based on multi-sensor continuous-discrete PMBM filter

A Single Measurement PHD Filter with Unknown Target Birth Intensity

Convolutional Recurrent Predictor: Implicit Representation for Multi-target Filtering and Tracking

A Robust TCPHD Filter for Multi-Sensor Multitarget Tracking Based on a Gaussian–Student's t-Mixture Model

Tracking A Large Number of Closely Spaced Objects Based on the Particle Probability Hypothesis Density Filter Via Optical Sensor

A Kernel Particle Probability Hypothesis Density Filter for Multi-target Tracking