Improved Automated Whale Signature Detection with Combined PAN and MSI Feature Vectors

Abstract

Estimation of improvement to automated whale signature detection using panchromatic and multi-spectral images when the two modalities area combined into a single feature vector. The overall goal of the current tasks on this project were to start to move to development of a useful detection capability that could be used operationally on WorldView imagery. What this meant was essentially to start working on mitigating false alarms in imagery that was not included in the training set so that we had a useful tool for extracting whale signatures from new imagery. We were able to develop new detection capabilities that now can run on an entire WorldView Ortho or Basic image and generate a much smaller number of detections that can easily be manually reviewed yet still maintained a high probably of detecting actual whale signatures. We now have a good start to an actual capability for a user to realistically review imagery to find new whale signatures.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Aug 06, 2023
Accession Number
AD1207443

Entities

People

  • Christopher Wackerman

Organizations

  • United States Naval Research Laboratory

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Algorithms
  • Classification
  • Data Science
  • Data Sets
  • Databases
  • Department Of Defense
  • Detection
  • Detectors
  • Eigenvalues
  • Eigenvectors
  • False Alarms
  • Gaussian Distributions
  • Histograms
  • Information Operations
  • Information Science
  • Military Research
  • Probability
  • Statistics
  • Training
  • Warning Systems

Fields of Study

  • Computer science

Readers

  • Computer Vision.
  • Marine Mammal Biology
  • Systems Analysis and Design