2019_programme: DEEP LEARNING BASED TECHNIQUE FOR ENHANCED SONAR IMAGING

Back


  • Session: 25. Signal and image processing
    Organiser(s): N/A
  • Lecture: DEEP LEARNING BASED TECHNIQUE FOR ENHANCED SONAR IMAGING
    Paper ID: 910
    Author(s): Rixon Fuchs Louise, Larsson Christer, Gällström Andreas
    Presenter: Rixon Fuchs Louise
    Presentation type: oral
    Abstract: Several beamforming techniques can be used to enhance the resolution of sonar\nimages. Beamforming techniques can be divided into two types: data independent\nbeamforming such as the delay-sum-beamformer, and data-dependent methods known as\nadaptive beamformers. Adaptive beamformers can often achieve higher resolution, but are\nmore sensitive to errors. Several signals are processed from several consecutive pings. The\nsignals are added coherently to achieve the same effect as having a longer array in synthetic\naperture sonar (SAS). In general it can be said that a longer array gives a higher image\nresolution. SAS processing typically requires high navigation accuracy, and physical array-\noverlap between pings. This restriction on displacement between pings limits the area\ncoverage rate for the vehicle carrying the SAS. We investigate the possibility to enhance\nsonar images from one ping measurements in this paper. This is done by using state-of-the art\ntechniques from Image-to-Image translation, namely the conditional generative adversarial\nnetwork (cGAN) Pix2Pix. The cGAN learns a mapping from an input to output image as well\nas a loss function to train the mapping. We test our concept by training a cGAN on simulated\ndata, going from a short array (low resolution) to a longer array (high resolution). The\nmethod is evaluated using measured SAS-data collected by Saab with the experimental\nplatform Sapphires in freshwater Lake Vättern
      Download the full paper
  • Corresponding author: Ms Rixon Fuchs Louise
    Affiliation: KTH
    Country: Sweden
    e-mail:

Back