Abstract

This thesis develops a deep learning based wound detection model (WUDEMO)that can detect which pixels in an image are part of a wound. WUDEMO will be used bya robot being built at Hofstra University. The purpose of the robot is to seal wounds using magnets. WUDEMO will allow the robot to detect the edges of the wound on which it will be attempting to place the magnets. This will allow the robot to know whereto place and glue the magnets to the skin. WUDEMO uses an artificial neural network as a convolution mask to detect which pixels are, or are not, part of a wound. WUDEMO was trained on a data set of 22 images, validated on three images, and subsequently tested on three completely separate images

Details

Title
Vision-Based Wound Detection System
Author
Kreidberg, Brian
Publication year
2022
Publisher
ProQuest Dissertations & Theses
ISBN
9798837534997
Source type
Dissertation or Thesis
Language of publication
English
ProQuest document ID
2704864959
Copyright
Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works.