Research on Ship Tracking Using Panorama Images - Towards River Traffic

In this research, the authors aim to detect ships sailing around the study ship from river navigation images using Faster R-CNN and predict their course. First, the authors cut out the part of the ship from the learning image and omit the image resizing process of Faster R-CNN. Next, ship detection experiments were carried out from river navigation images by a method for increasing the accuracy of ship detection using time series information. Finally, the authors reported a method for predicting the path of ships from the position and size of the rectangle containing ships detected after the ship position correction between image frames. The results are summarized as follows:(1) A data set was created by preprocessing the panoramic images created from river navigation images and the images taken by digital cameras. This dataset makes it possible to omit Faster R-CNN resizing processing and improve processing speed. (2)The minimum detection rate for ships sailing in rivers was about 40%.(3)As a result of course prediction, the error angle was 50 ° or less.

Language

  • English
  • Japanese

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01768380
  • Record Type: Publication
  • Source Agency: Japan Science and Technology Agency (JST)
  • Files: TRIS, JSTAGE
  • Created Date: Dec 24 2020 3:23PM