Presentation + Paper
5 June 2024 Multipolarization laser image fusion for improved underwater object recognition
Author Affiliations +
Abstract
Active illumination with underwater laser imaging has unique advantages for the identification of underwater objects, especially in shallow waters, complex marine environments and inaccessible locations. However, backscattered light from the water particulates can blur the resulting laser images. To improve the quality of underwater laser images, we have examined a wide range of image enhancement (IE) and restoration (IR) techniques. In our recent prior work, we have experimentally evaluated the efficacy of over 20 IE/IR methods specifically for the underwater object recognition, examining the impact of artifacts introduced by IE/IR on the deep neural network (DNN) architecture required for optimal classification accuracy. This paper builds on this work by considering the effect of polarization on underwater image restoration and object recognition. Using a one-of-a-kind multi-polarization underwater laser image dataset, this paper examines the image of polarization on the efficacy of IE/IR algorithms and proposes a deep neural network (DNN) for fusing and jointly exploiting the multi-polarization data for improved underwater object recognition.
Conference Presentation
(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.
Oladipupo O. Adeoluwa, Karsten Schnier, Anirban Swakshar, Seongsin M. Kim, Patrick Kung, and Sevgi Z. Gurbuz "Multipolarization laser image fusion for improved underwater object recognition", Proc. SPIE 13049, Laser Radar Technology and Applications XXIX, 130490D (5 June 2024); https://doi.org/10.1117/12.3014068
Advertisement
Advertisement
RIGHTS & PERMISSIONS
Get copyright permission  Get copyright permission on Copyright Marketplace
KEYWORDS
Polarization

Image fusion

Image enhancement

Single photon avalanche diodes

Sensors

Image processing

Object recognition

Back to Top