Vision-Based UAV Maritime Search and Rescue Using Point Target Detection

Abstract

Human maritime search and rescue missions have always been challenging
and an element of chance is involved in the detection of survivors at sea. This
research is proposing the use of machine vision to assist UAVs to increase the
chances of success in locating humans lost at sea.

This paper presents an application of current image processing methods for target
detection in a synthetic maritime scenario. An evaluation of the algorithm’s
performance is also provided. The difficulties faced in the automatic detection of
human targets in a maritime search environment are also considered.

The paper concludes that there is a range of greyscale intensities, approximately 26%
based on current data set, where the target was unable to be detected which may limit
the applicability of the algorithm. The effect on performance of target intensity level,
threshold and forgetting factor are also investigated.