Information Fusion for Industrial Mobile Platform Safety via Track-Before-Detect Labeled Multi-Bernoulli Filter

Abstract

This paper presents a novel Track-Before-Detect (TBD) Labeled Multi-Bernoulli (LMB) filter tailored for industrial mobile platform safety applications. At the core of the developed solution is two techniques for fusion of color and edge information in visual tracking. We derive an application specific separable likelihood function that captures the geometric shape of the human targets wearing safety vests. We use a novel geometric shape likelihood along with a color likelihood to devise two Bayesian updates steps which fuse shape and color related information. One approach is sequential and the other is based on weighted Kullback–Leibler average (KLA). Experimental results show that the KLA based fusion variant of the proposed algorithm outperforms both the sequential update based variant and a state-of-art method in terms of the performance metrics commonly used in computer vision literature.

Fast Track Smoothing Algorithm for Underwater Target Tracking Based on KD Tree

“…The labeled multi-Bernoulli algorithm is used in this paper. However, it is known that the labeled Bernoulli distribution family is not a conjugate prior for point trace measurements, and that the Generalized Labeled Multi-Bernoulli (GLMB) distribution family has been shown to be a conjugate prior [8], so we can convert the LMB form to the GLMB form after the prediction, and back to the LMB form after the update using the GLMB form [11]. It is equivalent to approximating the generalized labeled multi-Bernoulli filter by using the labeled multi-Bernoulli filter.…”

“…Then the labeled multi-Bernoulli random finite set ψ can be represented by the parameter set α ψ →  is defined. Then the density of the label multi-Bernoulli random finite set can be expressed as (11). Where ( ) l Y X is an inclusive function, see (10).…”

“…The smoothing algorithm based on labeled multi-Bernoulli [8][9][10][11] (LMB) is studied in this paper, and it is applied to the detection and tracking algorithm based on MUSIC, which is commonly used in underwater acoustics research. Finally, the algorithm simulation and performance comparison are performed.…”

On-Line Visual Tracking with Occlusion Handling

“…RFS multi-target filtering techniques such as Gaussian mixture and particle probability hypothesis density filters [26,40,41] have been applied to tracking from video data via detection in [42][43][44]. The more recent RFS based tracking algorithms such as multi-Bernoulli filter [39,45], labeled multi-Bernoulli (LMB) filter [46] and the Generalized labeled multi-Bernoulli (GLMB) filter [47,48] have been applied extensively in multi-object tracking with promising results [25,27,[49][50][51][52][53].…”

GLMB Tracker with Partial Smoothing

“…In particular, RFS forms the mathematical basis of many modern multi-object filters such as Probability Hypothesis Density (PHD) filter [3,4,5,6,7], cardinalized PHD (CPHD) filter [8,9,10], multi-Bernoulli filter [11,12], the Generalized Labeled Multi-Bernoulli (GLMB) filter [13,14,15,16,17,18,19], and its approximation the Labeled Multi-Bernoulli (LMB) filter [20,21]. In many applications, tracking algorithms rely on the standard point measurements to update the object states; in contrast, TBD [22,23,24,25] is an alternative approach that bypasses the detection module to directly exploit the observed spatial data. This technique is introduced under the RFS framework in Reference [26] with the development of the so-called separable likelihood model and, recently, in a hybrid (combination of standard observation and separable observation models) approach in Reference [27].…”