Multimodal Analysis of Drone based Imagery for Automatic Target Recognition

Abstract

This research proposes a large-scale dataset for training machine learning methods, a semantic segmentation network for visible and thermal images, and a novel annotation tool and data augmentation framework. We propose to perform semantic segmentation of the multi-modal imagery gathered from unmanned aerial vehicle platform, provided by our collaborators at US Army Research Laboratory. Our methodology makes use of both visible and longwave infrared imagery and provides per pixel semantic segmentation. Our proposed research will focus on i) novel annotation of training data, ii) novel transformations of the training data to be performed/used due to lack of large volumes of annotated data for our problem, and iii) robust and accurate semantic segmentation and matching of visible and LWIR imagery.

Document Details

Document Type
DoD Grant Award
Publication Date
Jun 17, 2020
Source ID
N000142012462

Entities

People

  • Chandra Kambhamettu

Organizations

  • Office of Naval Research
  • United States Navy
  • University of Delaware

Tags

Fields of Study

  • Computer science

Readers

  • Computer Vision.
  • Image Processing and Computer Vision.
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks
  • Autonomy