Real-Time Inland CCTV Ship Tracking

Lei Xiao,Zhongyi Hu,Minghai Xu

doi:10.1155/2018/1205210

Lei Xiao, Zhongyi Hu + Show 1 more

Open Access

https://doi.org/10.1155/2018/1205210

Copy DOI

Abstract

The predator algorithm is a representative pioneering work that achieves state-of-the-art performance on several popular visual tracking benchmarks and with great success when commercially applied to real-time face tracking in long-term unconstrained videos. However, there are two major drawbacks of predator algorithm when applied to inland CCTV (closed-circuit television) ship tracking. First, the LK short-term tracker within predator algorithm easily tends to drift if the target ship suffers partial or even full occlusion, mainly because the corner-points-like features employed by LK tracker are very sensitive to occlusion appearance change. Second, the cascaded detector within the predator algorithm searches for candidate objects in a predefined scale set, usually including 3-5 elements, which hampers the tracker to adapt to the potential diverse scale variations of the target ship. In this paper, we design a random projection based short-term tracker which can dramatically ease the tracking drift when the ship is under occlusion. Furthermore, a forward-backward feedback mechanism is proposed to estimate the scale variation between two consecutive frames. We prove that these two strategies gain significant improvements over the predator algorithm and also show that the proposed method outperforms several other state-of-the-art trackers.

Highlights

In the past decades, there are rapid progress in the domain of inland ship surveillance and management [1, 2]
According to compression sensing theory, if we want to get rid of the redundancy of the feature representations x1, x2 while maintaining discriminative power, the random projection matrix R should meet the demand of RIP [20], i.e., (1 − ε)‖x1−x2‖2l2 random Gaussian
When the target ship moves toward to the CCTV camera, the filter bank has the suitable filter to account for scale decrease

Summary

Introduction

There are rapid progress in the domain of inland ship surveillance and management [1, 2]. The extensive results showed that MIL tracker could greatly ease the sample ambiguity problem with fewer parameter tweaks but turned to select indiscriminative features Toward this issue, in [12], an EMIL (enhanced multiple instance learning) tracker was proposed. By doing so, the dynamic optimization between trackers and detectors is not established and we sacrifice the desired real-time property for a practical system Motivated by this insight, Kalal et al [13] proposed a novel predator framework for long-term face tracking in unconfined video streams. We aim to construct a practical inland CCTV ship tracking system To achieve this goal, motivated by the discussions, we build our algorithm based on predator framework since it has demonstrated favourable accuracy with decent time consumption.

Predator Framework

Short-Term Compressive Tracker

Experiments

Findings

Conclusions