2019_programme: EVALUATING HUMAN/MACHINE INTERACTION FOR UNDERWATER THREAT DETECTION AND CLASSIFICATION: A CASE STUDY.

Back


  • Session: 23. Mine countermeasures
    Organiser(s): N/A
  • Lecture: EVALUATING HUMAN/MACHINE INTERACTION FOR UNDERWATER THREAT DETECTION AND CLASSIFICATION: A CASE STUDY.
    Paper ID: 796
    Author(s): Lopera Tellez Olga Lucia
    Presenter: Lopera Tellez Olga
    Presentation type: oral
    Abstract: Although automatic target recognition algorithms have been developed for and are applied in different remote-sensing imaging applications, object-recognition tasks in maritime mine countermeasures are still performed almost exclusively by human operators. There is still a long way to see these automatic algorithms being fully trusted to take the place of human analysts. To facilitate this process, some authors suggest the development of methods by which humans and computers can work in concert to achieve improved performance. In this paper, a total of 1241 images (with 847 detection opportunities) collected from 7 different sonar systems (SSS and SAS) during naval mine hunting and route surveillance operations (mine countermeasures trials) are analysed by four expert operators, two scientists (with more than ten years of experience in sonar imaging) and by two automatic detection and classification algorithms (Markov Chain Monte Carlo and Adaptive Boosting Decision Trees). A seafloor segmentation map based on lacunarity and representing how difficult or how benign the seafloor is for object-recognition is used as a new strategy in order to divide the database between operator and computer. The impact of different factors on the object-recognition performance using different strategies are analysed in this paper: working on images with different resolution, working with previously unseen images (i.e., not previously used for training operators or algorithms), working with two kinds of operators: end-users and scientists. Results demonstrate the utility of considering the human operator as an integral part of the automatic underwater object recognition process, and demonstrate how automated algorithms can extend and complement human performances.
      Download the full paper
  • Corresponding author: Dr Lopera Tellez Olga Lucia
    Affiliation: Royal Military Academy, Ministry of Defence, Belgium
    Country: Belgium
    e-mail:

Back