Robust Visual Ship Tracking with an Ensemble Framework via Multi-View Learning and Wavelet Filter.

Xinqiang Chen,Yanguo Huang,Huixing Chen,Wenhui Zhang,Yongsheng Yang,Huafeng Wu,Pengwen Xiong

doi:10.3390/s20030932

Abstract

Maritime surveillance videos provide crucial on-spot kinematic traffic information (traffic volume, ship speeds, headings, etc.) for varied traffic participants (maritime regulation departments, ship crew, ship owners, etc.) which greatly benefits automated maritime situational awareness and maritime safety improvement. Conventional models heavily rely on visual ship features for the purpose of tracking ships from maritime image sequences which may contain arbitrary tracking oscillations. To address this issue, we propose an ensemble ship tracking framework with a multi-view learning algorithm and wavelet filter model. First, the proposed model samples ship candidates with a particle filter following the sequential importance sampling rule. Second, we propose a multi-view learning algorithm to obtain raw ship tracking results in two steps: extracting a group of distinct ship contour relevant features (i.e., Laplacian of Gaussian, local binary pattern, Gabor filter, histogram of oriented gradient, and canny descriptors) and learning high-level intrinsic ship features by jointly exploiting underlying relationships shared by each type of ship contour features. Third, with the help of the wavelet filter, we performed a data quality control procedure to identify abnormal oscillations in the ship positions which were further corrected to generate the final ship tracking results. We demonstrate the proposed ship tracker’s performance on typical maritime traffic scenarios through four maritime surveillance videos.

Highlights

A smart ship is considered in the ship industry as having the advantages of less carbon emissions, lower risk for the ship crew at sea, higher traffic efficiency, larger cargo carriage capability, etc., and it has attracted much research attention in the maritime traffic community [1,2,3]
We present the ship tracking results with the three trackers in detail for video #1, and verify the model’s performance on video #2, #3, and #4
Our proposed model performances were presented in detail with the wavelet filter (WF) model implemented with the above wavelet basis

Summary

Introduction

A smart ship is considered in the ship industry as having the advantages of less carbon emissions, lower risk for the ship crew at sea, higher traffic efficiency, larger cargo carriage capability, etc., and it has attracted much research attention in the maritime traffic community [1,2,3]. For the purpose of helping in smart ship maritime navigational environments, varied ship tracking techniques and data sources are employed to obtain informative static and kinematic ship data from varied maritime sources. Ship tracking data from maritime surveillance videos provide straightforward spatial-temporal information (e.g., ship trajectory, ship speeds, ship moving directions) which greatly enriches the situational awareness capability of the smart ship and, further improve maritime. By noticing potentially risky ship behaviors from the ship tracking results, the smart ship can inform the risk-involved ships to take early action (e.g., maneuver ship engines) to avoid potential maritime accidents. Maritime traffic participants can track ship positions with the LRIT technique over large time intervals (usually every six hours) when the ship travels far away from coastal areas Previous studies mainly employed automatic identification systems (AIS) to track ships sailing in inland waterways [4,5,6], and several techniques (e.g., synthetic aperture radar (SAR), long-range identification and tracking (LRIT)) have been integrated to further enhance ship tracking accuracy [7,8,9,10].

Methods

Results

Conclusion