Detection of Multiple Maneuvering Extended Targets by Three-Dimensional Hough Transform and Multiple Hypothesis Tracking

Bo Yan,Na Xu,Sheng Yang,Guangmin Wang,L. P. Xu

doi:10.1109/access.2019.2923357

Bo Yan, Na Xu + Show 3 more

Open Access

https://doi.org/10.1109/access.2019.2923357

Copy DOI

Abstract

Existing extended target probability hypothesis density (ET-PHD) filters are insufficient in tracking weak extended targets. Hough transform-based track-before-detect methods are designed to detect the weak targets in a straight-line constant-velocity model. Therefore, this paper presents a novel method for detecting and tracking multiple maneuvering weak extended targets by a 3-dimensional Hough transform (3DHT) and multiple hypothesis tracking (MHT). The proposed method consists of two stages. In stage 1, the measurements in multiple scans are partitioned into overlapped time windows. The tracklets in each window can be detected by the 3DHT. In stage 2, the tracklets are associated to get the entire trajectories by the MHT. The tracklets of weak targets can be detected by the 3DHT in stage 1. Association in stage 2 is designed to detect maneuvering targets. Some false alarm tracklets could be built in stage 1. However, the false alarm tracklets are independent and unlikely to form a sequential trajectory in stage 2. Merely, the trajectories whose target likelihood ratio larger than a detection threshold can be confirmed as a target. Both the real data and the synthetic data are performed with the proposed approach and several existing algorithms. The result infers that the proposed approach is superior to the others with much less prior information that is necessary.

Highlights

The false alarm tracklets are independent and unlikely to form a sequential trajectory in stage 2
Multiple target tracking is an essential requirement for surveillance systems
SYNTHETIC DATA extensive experiments are conducted to verify the feasibility of the proposed model from different aspects including robustness against various scenes, robustness to measurement noise, the ability of background suppression and target enhancement, target detection ability, and the computation time of the algorithm

Summary

INTRODUCTION

Multiple target tracking is an essential requirement for surveillance systems. A fully automatic tracking algorithm must be able to deal with an unknown number of targets, unknown target initiation and termination times, false measurements and possibly time-varying target trajectory behavior [1]. B. Yan et al.: Detection of Multiple Maneuvering Extended Targets by 3DHT and MHT of Ronald Mahler [5], the measurements (points) are partitioned into sets before the iteration of the PHD filter. The 3DHT-TBD is able to perform target detection, data association, track initiation, and track maintenance at the same time in an intense clutter environment. A method to detect and track multiple closely distributed weak and maneuvering extended target in a cluttered environment is proposed in this work. The 3-dimensional Hough transformation [2] and multiple hypothesis tracking algorithm is utilized to ensure the ability to detect weak target and maneuvering target respectively. The tracklets of different windows are associated to form the entire trajectory of each target with the multiple hypothesis tracking algorithm.

PRELIMINARIES

PROBLEM STATEMENT

ASSOCIATING THE TRACKLETS BY MHT

COMPLEXITY OF THE 3DHT-MHT

CONCLUSION