Human sonar operators typically outperform traditional passive contact follower algorithms when analysing broadband waterfall displays. One reason for this is that the current approaches, which incorporate Kalman and Alpha-Beta filters, are susceptible to the noise that is in abundance in underwater environments. Furthermore, they struggle to accurately track crossing contacts and detect quiet contacts in the presence of much louder ones. Since the quieter contacts are of most interest, this is a crucial capability of a tracking system. To reduce the cognitive burden on the operators, Ultra would like you to develop a detection and tracking algorithm that is more accurate and robust than the traditional methods, whilst not increasing the false alarm rate.

By looking at the sonar images provided, can you detect, classify and label the ten different objects which can be found underwater? The provided dataset of forward-looking sonar (FLS) images contain marine debris which have been submerged in a water tank and are situated on the tank floor. There are 10 different marine debris objects which were placed in the water tank. Not all the objects are in every image and not every image has an object in. Thales would like you to produce a method which detects and classifies the objects in the FLS image dataset. A more complex additional challenge would be to also produce a method which outlines and labels the objects.

Covert transmission of information in the underwater environment is of interest both in the defence sector and more widely. In the military context, covert signals could be used for secure communication with underwater assets without disclosing the location of the source, while such signals could also have the advantage of reduced anthropogenic noise in the underwater environment. One proposed method for covert communication is to ‘hide’ communications in artificial signals that would be expected in the marine environment, e.g. marine mammal vocalisations. Sonardyne and Newcastle University would like you to synthesise information-containing signals based on marine mammal vocalisations, examples of which can be found in the  Watkins Marine Mammal Sound Database.