Target Identity-aware Network Flow for online multiple target tracking

Abstract

In this paper we show that multiple object tracking (MOT) can be formulated in a framework, where the detection and data-association are performed simultaneously. Our method allows us to overcome the confinements of data association based MOT approaches; where the performance is dependent on the object detection results provided at input level. At the core of our method lies structured learning which learns a model for each target and infers the best location of all targets simultaneously in a video clip. The inference of our structured learning is done through a new Target Identity-aware Network Flow (TINF), where each node in the network encodes the probability of each target identity belonging to that node. The proposed Lagrangian relaxation optimization finds the high quality solution to the network. During optimization a soft spatial constraint is enforced between the nodes of the graph which helps reducing the ambiguity caused by nearby targets with similar appearance in crowded scenarios. We show that automatically detecting and tracking targets in a single framework can help resolve the ambiguities due to frequent occlusion and heavy articulation of targets. Our experiments involve challenging yet distinct datasets and show that our method can achieve results better than the state-of-art.

Problem Analysis of Multiple Object Tracking System: A Critical Review

“…This video sequence is a good choice to find out effectiveness of a tracking sequence, spatially when considered hijacking or centralization problem is one of the major issues for tracking. It shows the recent two works that is target identity network flow [7] and subgraph decomposition methods shows [8] better results to track crossing people each others.…”

“…proposed a multiple object tracking system where targets identity based network flow method is used to track multiple objects [7]. Here structured learning and inference is used to detect an object.…”

“…Andriyenko Energy minimization method [3] 60.5% Caglayan Dicle Iterative Hankel Total Least Squares method [5] 88.0% Brendels maximum weight independent set method [6] 85.9% Afshin Dehghan's target identity network flow method [7] 92.9%…”

Image-Based Multi-Target Tracking Through Multi-Bernoulli Filtering with Interactive Likelihoods

“…In situations in which target detection itself is challenging, such as in scenarios with substantial clutter, track-before-detect methods tend to perform significantly better [55][56][57]. The contribution presented in this paper, hence, falls into the second category, track-before-detection, in which there are no scanning or searching schemes involved, and detections are not strictly necessary.…”

GMMCP tracker: Globally optimal Generalized Maximum Multi Clique problem for multiple object tracking

“…Different variations of network flow are also used in MOT recently [12,10,22]. Authors in [10] incorporate constant velocity motion model in network flow graph and proposed a Lagrangian relaxation solution to min-cost max-flow problem.…”