Social Force Model-Based MCMC-OCSVM Particle PHD Filter for Multiple Human Tracking

Abstract

Video-based multiple human tracking often involves several challenges, including target number variation, object occlusions, and noise corruption in sensor measurements. In this paper, we propose a novel method to address these challenges based on probability hypothesis density (PHD) filtering with a Markov chain Monte Carlo (MCMC) implementation. More specifically, a novel social force model (SFM) for describing the interaction between the targets is used to calculate the likelihood within the MCMC resampling step in the prediction step of the PHD filter, and a one class support vector machine (OCSVM) is then used in the update step to mitigate the noise in the measurements, where the SVM is trained with features from both color and oriented gradient histograms. The proposed method is evaluated and compared with state-of-the-art techniques using sequences from the CAVIAR, TUD, and PETS2009 datasets based on the mean Euclidean tracking error on each frame, the optimal subpattern assignment metric, and the multiple object tracking precision metric. The results show improved performance of the proposed method over the baseline algorithms, including the traditional particle PHD filtering method, the traditional SFM-based particle filtering method, multi-Bernoulli filtering, and an online-learningbased tracking method. Index Terms-Multiple human tracking, Markov chain Monte Carlo (MCMC) resampling, one class support vector machine (OCSVM), probable hypothesis density (PHD) filter, social force model. I. INTRODUCTIONV IDEO based multiple human tracking plays an important role in many applications such as surveillance, guidance, and homeland security, especially in enclosed environments such as an airport, campus or shopping mall. Tracking multiple human targets in the above situations presents several challenges including varying number of targets, object occlusion, and the adverse effect of environmental noise within measurements [1],

Adaptive Online Learning With Regularized Kernel for One-Class Classification

“…Finally, weight matrix for initial samples is obtained, which needs to be updated as follows: Initially, Φ and P are calculated as per (35). Both are updated continuously for any upcoming new samples X v as per (11)- (19). After processing the training dataset X tr curr in the current sliding window, output function for any set of k samples X k = {x 1 , x 2 ....x k } can be written as follows:…”

“…First, if any multiplication algorithm [73]. Second, Woodbury formula [62] is used to solve (18) as shown in (19). After using Woodbury [62] in (19), the inverse of two matrices, i.e.…”

“…SVM and KPCA. SVDD based incremental one-class classifier (incSVDD) for online learning has been developed by Tax and Laskov [22] and further explored by other researchers [18], [19]. KPCA based online one-class classifier (OKPCA) has been developed by Chatzigiannakis and Papavassiliou [23], where the model is constructed in a batch mode for all new samples.…”

Multi-Level Cooperative Fusion of GM-PHD Filters for Online Multiple Human Tracking

“…GM-PHD and Sequential Monte Carlo (SMC)-PHD filters are two commonly used implementations in this theory, as they have been able to generate convincing tracking performance in video-based multi-target tracking [2], [3], [5], [7], [15]- [17]. This is attributed to the advantages of PHD filtering methods, as they have the ability to deal with varying number of targets, and also provide the estimates in both cardinality and localization with relatively low computational cost [2]. However, conventional PHD filters are inherently unable to assign identity to targets, so an additional labelling mechanism is needed for completeness, such as the early association in [15] and the post-processing step in [17].…”

“…In this framework, they developed a bilinear Long Short-Term Memory (LSTM) which jointly encodes both appearance and motion information for target tracks, in order to make the full use of past target appearances. Since our proposed tracking algorithm focuses on the GM-PHD filtering framework, we also review several recently developed PHD filtering based trackers for multiple human tracking, such as a social force model aided particle PHD filter [2], SMC-PHD filter with online group-structured dictionary learning [5], early-association based particle PHD filter [15] and GM-PHD filter with hierarchical association [31]. These methods did not fully address the issues of identity labelling or target ambiguity in the PHD filter.…”