Learning an Image-Based Motion Context for Multiple People Tracking

Abstract

We present a novel method for multiple people tracking that leverages a generalized model for capturing interactions among individuals. At the core of our model lies a learned dictionary of interaction feature strings which capture relationships between the motions of targets. These feature strings, created from low-level image features, lead to a much richer representation of the physical interactions between targets compared to hand-specified social force models that previous works have introduced for tracking. One disadvantage of using social forces is that all pedestrians must be detected in order for the forces to be applied, while our method is able to encode the effect of undetected targets, making the tracker more robust to partial occlusions. The interaction feature strings are used in a Random Forest framework to track targets according to the features surrounding them. Results on six publicly available sequences show that our method outperforms state-of-theart approaches in multiple people tracking.

Discovering Groups of People in Images

“…Using short tracks extracted from video data they jointly cluster people into groups and derive their longer trajectories. Leal-Taixe et al [22] extend this idea further to incorporate more generalized interactions between people.…”

Bayesian Multi-object Tracking Using Motion Context from Multiple Objects

“…In [24], the pairwise relative object motion model is developed as an additional similarity function for MOT, which uses both past and future tracklets. Leal-Taixé et al [11] propose robust motion models which are trained in offline manner based on motion training samples. On the contrary, our method does not need training data for constructing motion models.…”

Improving Multi-frame Data Association with Sparse Representations for Robust Near-online Multi-object Tracking

“…In some approaches, the multi-frame data association problem has been formulated in a more specific class of problems, like for example minimum cost flow problems [5][6][7]12], binary integer programming [14], maximum weighted clique [13] or independent set [15]. The main advantage of such approaches is that efficient optimization methods designed for these problems can be directly employed to find the data association solution.…”

“…Among offline techniques, global approaches perform the data association over all the frames simultaneously or by batch [5][6][7][8][9][10][11][12][13][14][15], whereas sliding window (a.k.a. multi-scan, near-online, or online with delay) methods optimize only a few recent frames at the same time [16][17][18][19][20].…”