RT Journal Article
A1 P. Sévigny
AD Defence Research and Development, Ottawa, ON, Canada
A1 J. Fournier
AD Defence Research and Development, Ottawa, ON, Canada

PB iet
T1 Automated stationary human target detector for 3D through-wall radar imagery
JN Electronics Letters
VO 53
IS 15
SP 987
OP 991
AB The use of through-wall radar imagery for remote intelligence of building interiors is promising but challenging. Relevant information about human targets and room layout features is typically buried in clutter. In this Letter, the authors propose a methodology for automated extraction of information about stationary human targets behind walls. Based on treatment of the individual blobs found in the imagery, the method consists of thresholding, segmentation, classification and 3D visualisation. Although further optimisation of each of the steps is required, they demonstrate with two examples, including one cluttered scene, that the combination of these steps is effective at eliminating large amounts of clutter and identifying human targets behind walls.
K1 intermediate labelled images
K1 automated stationary human target detector
K1 3D through-wall radar imagery
K1 local majority filter
K1 random decision forest model
K1 pixel labelling
K1 3D visualisation
K1 image thresholding
K1 statistical perspective
K1 saliency map generation
K1 automated information extraction
K1 voting algorithm
K1 real-time systems
K1 local majority saliency map
K1 decision redundancy
K1 ensemble machine learning schemes
K1 image classification
K1 image segmentation
K1 voting pool
K1 uncertainty challenges
K1 enhanced decision fusion
K1 semantically segmented images
K1 building interior remote intelligence
DO https://doi.org/10.1049/el.2017.1454
UL https://digital-library.theiet.org/;jsessionid=bhf33dkhbie35.x-iet-live-01content/journals/10.1049/el.2017.1454
LA English
SN 0013-5194
YR 2017
OL EN