Abstract
An enhanced sequential Monte Carlo probability hypothesis density (PHD) filter-based multiple human tracking system is presented. The proposed system mainly exploits two concepts: a novel adaptive gating technique and an online group-structured dictionary learning strategy. Conventional PHD filtering methods preset the target birth intensity and the gating threshold for selecting real observations for the PHD update. This often yields inefficiency in false positives and missed detections in a cluttered environment. To address this issue, a measurement-driven mechanism based on a novel adaptive gating method is proposed to adaptively update the gating sizes. This yields an accurate approach to discriminate between survival and residual measurements by reducing the clutter inferences. In addition, online group-structured dictionary learning with a maximum voting method is used to robustly estimate the target birth intensity. It enables the new-born targets to be automatically detected from noisy sensor measurements. To improve the adaptability of our group-structured dictionary to appearance and illumination changes, we employ the simultaneous code word optimization algorithm for the dictionary update stage. Experimental results demonstrate our proposed method achieves the best performance amongst state-of-the-art random finite set-based methods, and the second best online tracker ranked on the leaderboard of latest MOT17 challenge.
Highlights
Video-based multiple human tracking has been an emerging technique in the last decade, since it is crucial in many applications such as intelligent video surveillance, behavior analysis, assistive technology and human-computer interactions [2]–[4]
Wu et al [24] improved the GM-probability hypothesis density (PHD) filter combined with an iterative random sample consensus (I-RANSAC) method to estimate the target birth intensity from uncertain measurements, whereas they approximated the trajectory of a newborn target as a straight line by means of regressing a line model with a given measurement set, which is not always feasible to track targets with nonlinear movements especially in video surveillance
These three trackers are all reliant on the random finite set (RFS)-based Bayesian filtering method, including conventional particle PHD filter (SMC-PHD) [21], background subtraction based multiBernoulli filter (MB) [26], and social force model based particle PHD filter (SFM-PHD) [2]
Summary
Video-based multiple human tracking has been an emerging technique in the last decade, since it is crucial in many applications such as intelligent video surveillance, behavior analysis, assistive technology and human-computer interactions [2]–[4]. Due to the imperfections in the human detector, two challenging issues in the SMC-PHD filter remain that are accurately discriminating the survival and birth measurements from the original detection results with uncertainty, as well as adaptively determining the birth intensity of the newborn targets The multi-task group-structured sparsity is achieved by exploiting a collaborative hierarchical Lasso (C-HiLasso) model [27] to strengthen the discriminability of the sparse coefficients at the group level In this way, a maximum voting method based on the sparsity solution is proposed to eliminate the existing interferences induced by noise or clutter from the measurement set, which leads to increasing the accuracy of birth intensity generalization.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.