Maritime Search and Rescue Missions with Aerial Images: A Survey

TL;DR

This survey addresses the problem of detecting people at sea from aerial imagery for maritime search and rescue. It surveys methods spanning classification, segmentation, detection, and tracking, with emphasis on deep learning, attention mechanisms, and synthetic data to overcome scarce real-world data. The authors synthesize datasets and evaluation metrics, compare state-of-the-art results across multiple benchmarks, and discuss practical trade-offs between accuracy and on-board efficiency. They identify key challenges—data scarcity, domain shift, tiny targets, and environmental variability—and highlight synthetic data, domain adaptation, and lightweight architectures as pivotal for real-world deployment and future research directions.

Abstract

The speed of response by search and rescue teams at sea is of vital importance, as survival may depend on it. Recent technological advancements have led to the development of more efficient systems for locating individuals involved in a maritime incident, such as the use of Unmanned Aerial Vehicles (UAVs) equipped with cameras and other integrated sensors. Over the past decade, several researchers have contributed to the development of automatic systems capable of detecting people using aerial images, particularly by leveraging the advantages of deep learning. In this article, we provide a comprehensive review of the existing literature on this topic. We analyze the methods proposed to date, including both traditional techniques and more advanced approaches based on machine learning and neural networks. Additionally, we take into account the use of synthetic data to cover a wider range of scenarios without the need to deploy a team to collect data, which is one of the major obstacles for these systems. Overall, this paper situates the reader in the field of detecting people at sea using aerial images by quickly identifying the most suitable methodology for each scenario, as well as providing an in-depth discussion and direction for future trends.

Figures (12)

• Figure 1: Graphical overview of the task, concerns, UAV equipment, and background limitations.
• Figure 2: General scheme for the classification task. First, the images are passed through a feature extractor, and then a classifier is used to decide to which category each image belongs.
• Figure 3: Examples of the different perspectives for segmentation tasks for maritime rescue operations. In this particular case, the images present objects belonging to two categories: person and rock.
• Figure 4: Example of the segmentation approach by Mendonça et al. mendoncca2016cooperative. The image on the left depicts the aerial image of a swimmer, the image on the center shows the saliency-based target detection, and the image on the right shows the final mask after applying the binary threshold filter.
• Figure 5: General pipelines for learning-based object detection methods, illustrating both one-stage and two-stage architectures. Image extracted from kang2022survey.