Maritime situational awareness using adaptive multi-sensor management under hazy conditions

Dilip K. Prasad, Chandrashekar Krishna Prasath, Deepu Rajan, Lily Rachmawati, Eshan Rajabally, Chai QuekarXiv 2017 · arXiv

Abstract

Autonomous maritime vessels (AMVs) require reliable situational awareness to navigate safely, especially in regions prone to poor visibility like the Southeast Asian haze. This paper proposes a robust multi-sensor architecture that adaptively manages inputs from visible light, infrared, radar, and weather sensors. By intelligently weighting these data sources based on environmental conditions, the system maintains a clear operational picture even when traditional electro-optical sensors are degraded.

Introduction

Navigating a massive ship is a complex task; doing so without a human at the helm in thick haze is nearly impossible with current standard technology. While sensors like radar are excellent for long-range detection, they have "shadow regions" or blind spots close to the vessel. Conversely, standard electro-optical (EO) visible-range cameras are severely degraded by haze, fog, and night conditions.

The authors address the challenge of "all-weather" autonomy by arguing that the primary bottleneck is not sensor hardware, which is already mature, but rather the intelligent integration and analysis of multi-modal data.

Methodology: Adaptive Multi-Sensor Management

The proposed architecture centers on an Adaptive Multi-Sensor Management block that serves as the system's primary intelligence.

Figure 1: The proposed adaptive multi-sensor management architecture integrates diverse data sources to maintain situational awareness.

Key components of this architecture include:

  • Adaptive Control Block: This unit monitors non-imaging data, such as humidity and aerosol levels, to determine the optimal processing strategy for imaging data. For instance, in heavy haze, it automatically increases the importance of Infrared (IR) sensors over visible-light cameras.
  • Sensor Fusion: The system integrates a diverse array of inputs, including imaging (Radar, Sonar, Visible EO, and Infrared), weather/geospatial data (Rain sensors, GPS), and external data (AIS and VTS).
  • Computational Intelligence: Utilizing fuzzy logic, this block resolves conflicts between sensors—such as when radar detects an object that a camera cannot see—and identifies high-risk navigation scenarios.
  • Markov Random Processes: Weather states are modeled as a connected network, allowing the system to predict transitions (e.g., from clear to hazy) and adjust sensor sampling frequencies accordingly.

The NIR Advantage in Hazy Conditions

A significant insight from this research is the visual proof that Near-Infrared (NIR) imaging can "see through" haze that completely obscures ships in the visible spectrum. This phenomenon occurs because the longer wavelengths of IR light are less prone to scattering by haze particles.

Figure 2: Qualitative comparison showing the visibility delta between Near-Infrared (NIR) and Visible (VIS) spectrums in hazy conditions (Singapore, PSI: 230).

In conditions with approximately 5km visibility, the authors demonstrate that NIR sensors can detect multiple vessels that are entirely invisible to standard VIS cameras. This makes EO sensors a safety requirement rather than an option, as they fill the critical near-range gap left by radar.

Conclusion

The paper presents a conceptual but highly practical architecture for maritime autonomy. By treating weather data as a control signal rather than just an environmental condition, the system can dynamically adjust its sensor suite. While the seamless, real-time integration required for zero-human-intervention vessels is still undergoing development, this hierarchical weighting approach—prioritizing AIS/VTS and using a distance-based handoff between EO and radar—provides a clear path forward for all-weather maritime navigation.

Summary updated

BibTeX
@misc{prasad2017maritime,
  author = {Dilip K. Prasad and Chandrashekar Krishna Prasath and Deepu Rajan and Lily Rachmawati and Eshan Rajabally and Chai Quek},
  title = {{Maritime situational awareness using adaptive multi-sensor management under hazy conditions}},
  year = {2017},
  eprint = {1702.00754},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1702.00754},
}
Figure