A tracking algorithm suitable for embedded systems implementation

Abstract

Particle filters have been widely used for video tracking due to their robustness. However, most particle filter algorithm implementations are computationally expensive which makes them ill-suited for real-time embedded systems. There have been some attempts to provide hardware implementations for the particle filter, but none of them tried to simplify the algorithm first in order to make it more efficient for the hardware implementation. In this paper, a new sampling algorithm inspired from the particle filter methodology is proposed. It includes a resampling scheme that uses a new method to assign the number of particles between filter iterations and a criterion to reduce the number of processed samples, both in order to reduce the computational burden. Our experiments demonstrate that the algorithm can be as accurate as the CONDENSATION algorithm, while reducing the computational load by a factor of 30%.I.

A reformulated systematic resampling algorithm for particle filters and its parallel implementation in an application-specific instruction-set processor

“…They approximate the a posteriori probability density function of an unknown state using a group of particles of significant weight. Due to their outstanding accuracy compared to Extended Kalman Filters [3] for non-linear and/or non-Gaussian applications, they have gained in popularity in many practical applications such as visual tracking [4] [5] and navigation [6] [7].…”

Parallel array histogram architecture for embedded implementations

“…Their calculation is a relatively simple operation and it is often performed by general-purpose processors (GPPs). Nonetheless, hardware acceleration may be necessary to satisfy high throughput requirements for some real-time applications, such as histogram-based particle filters for video object tracking [1,2].…”