A multi-frame particle tracking algorithm robust against input noise

Abstract

The performance of a particle tracking algorithm which detects particle trajectories from discretely recorded particle positions could be substantially hindered by the input noise. In this paper, a particle tracking algorithm is developed which is robust against input noise. This algorithm employs the regression method instead of the extrapolation method usually employed by existing algorithms to predict future particle positions. If a trajectory cannot be linked to a particle at a frame, the algorithm can still proceed by trying to find a candidate at the next frame. The connectivity of tracked trajectories is inspected to remove the false ones. The algorithm is validated with synthetic data. The result shows that the algorithm is superior to traditional algorithms in the aspect of tracking long trajectories.

Higher order multi-frame particle tracking velocimetry

“…Oulette et al (2006) used criteria as the minimal acceleration for the third frame, or minimized change in acceleration for the fourth frame, where a modified version of the latter criteria showed the best results. Li et al (2008) developed a technique using information of previous five frames to determine the particle image position in the sixth frame. Due to the large number of previous frames, their algorithm is very robust to noise and a method was developed to gap even frames with missing particle information.…”

Volumetric Monitoring and Modeling of Indoor Air Pollutant Dispersion by the Use of 3D Particle Tracking Velocimetry

“…• "Correct tracking" ratio 2D (respectively 3D): Proposed by (Li 2008), it is the number of 2D (respectively 3D) tracked positions which are identical to the actual 2D/3D particle positions divided by the total number of tracked positions. This ratio only deals with tracked trajectories and is an indicator of the tracking accuracy.…”

“…Particle tracking schemes can be divided into three main categories: -Image plane based tracking schemes: Particles are tracked on each camera 2D image plane separately through time (temporal tracking). Afterwards, the resulting 2D trajectories are matched in 3D object space and the 3D coordinates are calculated (Biwole et al, 2009;Engelmann, 1998Engelmann, , 2000Ge & Cha, 2000;Guenzennec et al, 1994;Jähne, 1997;Li et al 2008;Wierzimok & Hering, 1993) -Object plan based tracking schemes: Particle 3D coordinates are first calculated at each time step separately. Afterwards, the resulting set of time-ordered 3D coordinates is the only input for temporal tracking directly in object space.…”

“…Based on (Li et al, 2008) 2D tracking scheme, (Biwole et al, 2009) algorithm first utilizes a second order polynomial regression method to predict the center of the search region. The purpose of the regression is to minimize the changes in particle acceleration:…”

“…As will be shown in section 5, 3D matching is later checked several times throughout the trajectory. A more detailed description of this 2D tracking scheme is given in (Li et al, 2008). The scheme was validated using simulated data of a jet flow impinging on a wall.…”