Autoencoded Reduced Clusters for Anomaly Detection Enrichment (ARCADE) in Hyperspectral Imagery

Abstract

Anomaly detection in hyper-spectral imagery is a relatively recent and important research area. The shear amount of data available in a many hyper-spectral images makes the utilization of multivariate statistical methods and artificial neural networks ideal for this analysis. Using HYDICE sensor hyperspectral images, we examine a variety of pre-processing techniques within a framework that allows for changing parameter settings and varying the methodological order of operations in order to enhance detection of anomalies within image data. By examining a variety of different options, we are able to gain significant insight into what makes anomaly detection viable for these images, as well as what impact parameter and methodology changes can have on the total classification effectiveness, false positive fraction and true positive fraction regarding classification.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 24, 2016
Accession Number
AD1053993

Entities

People

  • Brenden A Mclean

Organizations

  • Air Force Institute of Technology

Tags

Communities of Interest

  • Energy and Power Technologies
  • Sensors

DTIC Thesaurus Topics

  • Air Force
  • Anomaly Detection
  • Change Detection
  • Classification
  • Data Mining
  • Data Science
  • Detection
  • Detectors
  • Dimensionality Reduction
  • Electromagnetic Spectra
  • Factor Analysis
  • Feature Extraction
  • Hyperspectral Imagery
  • Information Processing
  • Information Science
  • Neural Networks
  • Two Dimensional

Fields of Study

  • Computer science

Readers

  • Distributed Systems and Data Platform Development
  • Image Processing and Computer Vision.
  • Organizational Psychology.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks