ROBUST H2020 project
  • Home
  • Objectives
  • Work Plan
    • WP 1: Seabed mining and operational requirements of AUV.
    • WP2: LIBS development
    • WP3: Automatic recognition of mining targets
    • WP4: Sea bed mapping and volume measurement
    • WP5: AUV and manipulator development
    • WP6: Control architecture
    • WP7: Manufacturing, integration and preliminary testing
    • WP8: Large scale trials and LCCA
    • WP9: Dissemination and exploitation
    • WP 10: Management and Coordination
    • WP 11: Environmental Protection Question Requirement.
  • Partners
  • News
  • Publications
  • Contact us
  • Files
  • Intranet
  • Privacy

WP3: Automatic recognition of mining targets

  • Home
  • WP3: Automatic recognition of mining targets

WP3 “Automatic Recognition of Mining Targets” focuses on technologies for automatic identification of mining targets in acoustic data, oriented towards general mapping of the survey area and automatic detection and localization of targets using optical information for close range studies.

For acoustic detection of Mn-nodules a multibeam echo sounder is used. Both, the bathymetry and backscatter provided by this device can contain valuable information. The proposed strategy aims to find the largest area of high Mn-nodule likelihood. Additional working constraints of the AUV are considered, too. Four different features are used for determining the promising Mn-nodule areas.

Digital elevation map (bathymetry) of a seafloor area
Slope
Topographic Position Index (TPI)

Grid obtained by combining multiple binary classifiers as outlined in the previous section. The map covers an area of about 4.6 km width and 1.7km length. The green parts of the map indicate the most promising regions.

The visual detection of Mn-nodules is based on a deep learning approach. The first step is to perform an adequate image pre-processing on the acquired images. This is necessary due to inhomogeneous illumination and other underwater effects like backscatter and attenuation. For these tasks we perform lighting inhomogeneity correction, noise reduction with bilateral filtering and image dehazing.

20150802_213000_IMG_3562.JPG
Orignal Images with lighting inhomofeneity and vignetting-like artifacts
Image Dehazing: CLAHE applied on the previous zoomed image

In this project, we compared the performance of Fast-RCNN with DarkNet Yolo. Both algorithms are general purpose object detection schemes with fast and accurate implementations. We obtained good results using both algorithms, due to the high frame rate that Yolo is able to rich, we will use it on the AUV for real time manganese nodule detection

Fully annotated image using DIAS system
Darket YOLO detection example after training. Green boxes indicate ground truth annotated nodules, red bounding boxes indicates detected nodules

Recent Posts

  • Pressure housing successful test up to 600 bar
  • LIBS system ready for its journey to the Pacific Ocean
  • Successful integration of the LIBS system into the pressure housing
  • LIBS Housing ready for new endeavors
  • LIBS Press Release

Archives

  • December 2018
  • November 2018
  • June 2018
BACK TO TOP

Social

Follow me on:

© Robust. All Rights Reserved.