Target detection using Gaussian mixture models and fourier transforms for UAV maritime search and rescue

In the event of a maritime disaster, casualties need to be found and rescued promptly. Image processing methods could help to perform automated detection from a UAV. The main current approaches make use of multispectral and thermal cameras, which can deal with lightning difficulties but are expensive and could suffer from high noise problems. This paper presents a method combining both color analysis and frequency patterns identification using an inexpensive vision camera (EO camera), and implements it through an adaptive algorithm to deal with a dynamically changing background. The method is tested successfully in different environments.