1)Deep Learning applied to Underwater Mine Warfare
In this article, we have presented our work on Underwater
Mine Detection in SAS images. We have setup a new pipeline of operation composed of (i) a synthetic image generator, (ii) a DAE, (iii) a background dataset extractor and (iv) a DCNN as illustrated on Fig. 1. Through our tests, we show that this pipeline is very promising as results on a first simple operational dataset are encouraging. However, work needs to be pursued in order to be able to tackle more complex and challenging backgrounds. We foresee a few tasks to be carried out on both the image generation part and the machine learning part.
2)A Study on Detection and Classification of
Underwater Mines Using Neural Networks
This paper has examined various neural network
techniques which have the potential to aid the detection and classification of mine-like objects in side scan sonar imagery. In side scan sonar-imaging applications, five components of the detection and classification system were examined. These components are Data preprocessing, Segmentation, Feature extraction, Detection and Classification. For each of these components, neural network techniques is used to improve the performance of underwater mine, and side scan sonar systems where discussed. Examples of successful or instructive methods from the literature were given. Finally, some general neural network considerations common to each imaging methodology were given. Table 1 present the selected overview of neural network techniques among the existing systems and also display the finding rate of mine detection and classification. The need of human element of the mine hunting system is emphasized.