Dynamic Environment Modeling with Gridmap: A Multiple-Object Tracking Application

Abstract

The Bayesian occupancy filter (BOF) [1] has achieved promising results in the object tracking applications. This paper presents a new development of BOF which inherits original BOF's advantages in handling occlusion and representing objects' shape. Meanwhile, the new formulation has significantly reduced original BOF's complexities and can be run in realtime. In Bayesian occupancy filter, the environment is finely divided into 2-dimensional grids. Different from conventional occupancy gridmaps, in BOF, each grid has both static (occupancy) and dynamic (velocity) characteristics. In the new proposed BOF, the velocity of each cell is modeled as a distribution. The distribution for each cell occupancy can therefore be inferred using a filtering mechanism. Like the original BOF, no representation of objects exists in the BOF gridmap. However, there are often applications which require the definition and tracking at the object level. In the post-processing, a segmentation algorithm is implemented to extract the objects from BOF estimation. Thereafter, standard target tracking methods are employed to further analyze each object's motion. Experiments using data from an indoor human tracking application demonstrate that our approach yields satisfactory results even when serious occlusions exist.

Conditional transition maps: Learning motion patterns in dynamic environments

“…Bayesian Occupancy Filters (BOF) [5], [6], [9] have been applied to object tracking. BOF model the environment using a grid representation, with additional velocity components.…”

Environment perception using dynamic polylines and particle based occupancy grids

“…By estimating the occupancy of each cell in the 4D grid, the speeds for the classical cells in the 2D grid can be computed. Another solution for the representation of speeds is presented by Chen et al, in [6]. Instead of having a 4D grid, this solution comes back to 2D, but uses for each cell a distribution of speeds, in the form of a histogram.…”

Grid mapping in dynamic road environments: Classification of dynamic cell hypothesis via tracking

“…A computationally less demanding modification [8] produces a two-dimensional occupancy grid representation with each cell having an associated distribution over possible velocities. Another approach is described in [9] where a particle filterbased tracking is used to infer the velocity of each grid cell.…”