Robust automatic ship tracking in harbours using active cameras

Abstract

Radar is commonly used to detect and track ships in maritime surveillance. Unfortunately the systems are costly and do not provide any visual information about the object's type. To complement the ship identity information given by a radar system, we propose a supplementary system using active visual cameras that can robustly detect and track ships in harbours. By combining a high-quality, non real-time robust object detector with a feature point tracker with low computational complexity, it is possible to track ships in real time over long intervals and large distances. In addition to controlling pan and tilt, we dynamically control camera zoom to provide a high resolution image of the tracked object over a large range of distances. The tracking system is improved by a special motion estimation model for the feature points, which also incorporates zooming of the camera. The system is robust and sustains tracking even under challenging conditions, such as multiple viewpoints, a large variety of ships and various weather conditions. During experiments, various types of ships were successfully tracked for up to 18 minutes, and over a distance of almost 1.5 km in the port of Rotterdam. The proposed system is generic and can be utilized in various tracking applications, by training the detector for a different object class.

Sea–Sky Line and Its Nearby Ships Detection Based on the Motion Attitude of Visible Light Sensors

“…Frost et al [24] also applied the prior knowledge of ship shape to the level set segmentation algorithm to improve ship detection results. Loomans et al [25] integrated a multi-scale Histogram of Oriented Gradient (HOG) detector and a hierarchical Kanade-Lucas-Tomasi (KLT) feature point tracker to track ships in the port, and achieved better detection and tracking effects. The above algorithm is not based on the background subtraction algorithm, but is based on the manually set ship characteristics for target detection.…”

Evaluating deep semantic segmentation networks for object detection in maritime surveillance

“…Recent approaches for maritime object detection [2,3,5,7,11,12,13,20,22,23] include object classification, background modelling, and saliency based methods. Detection based on classifiers using Haar [2], HOG [12] or CNN [3,7] features require substantial training data and are prone to overfitting on specific subsets of maritime objects. Modelling maritime backgrounds (i.e.…”

Temporally stable feature clusters for maritime object tracking in visible and thermal imagery

“…Some approaches apply learning-based detectors (e.g. HOG [9], MACH [13]) to capture general shape and appearance information about objects. However, the detector would have to be exposed to a formidable training effort to capture all possible variations of vessels that might be observed.…”

“…This includes a highly dynamic background whereby lighting and meteorological conditions severely influence the motion and appearance of waves, the variety of objects and their profiles (ranging from skiffs to fishing boats to oil tankers), varying object dynamics and appearance, and non-stationary sensors. Existing representative maritime trackers [5,6,7,8,9,13,14,15] are limited in that some rely on strong context, such as reliably detecting the horizon [6], require substantial training [9,13], operate on only a single modality (visible [6,8,9,13], thermal [5,14,15] with the exception of [7]), or are not robust to significant camera movement (for example, panning), or scale changes (including small targets) [7,16].…”