Artificial neural nets to detect lines in noise

This work discusses some investigations on the detection of lines in noise. The motivation comes from the effort being put into naval sonars to detect sources of decreasing signal strength. This has as a consequence an increase in the total information acquired by the detection process. The information needs assessing and sorting in real time, with much of it being discarded. To do this with current methods would require more operators but these are too expensive to be used in the numbers needed to extract all the information being collected by the sonar. Artificial neural nets (ANN) have a reputation for being able to cope with some problems that conventional signal processing finds difficult. In particular for this problem was ANN´s supposed ability to find patterns hidden in noise. This investigation´s aim was to validate or otherwise this ability for lofargrams. This problem requires the recognition of simple patterns (a line) on a noisy background