Machine vision for detection of invertebrates on crops /
Files
(Published version)
Date
2018
Authors
Liu, Huajian,
Editors
Advisors
Journal Title
Journal ISSN
Volume Title
Type:
thesis
Citation
Statement of Responsibility
Conference Name
Abstract
The aim of this research is to study the feasibility of using multispectral and three-dimensional (3D) vision systems to detect common invertebrates on crops. The unstable sunlight, camouflage of pests and complex morphological and spectral features of pests and plants raise many challenges. The spectral features of UV images were first studied and it was found that UV images can improve the accuracy of detection when fused with visible light and near-infrared (NIR) images. Then, inspired by the colour perception of birds and insects, a colour space named hyper-hue-saturation-intensity (HHSI) was derived. Hyper-hue is independent of saturation and intensity and more suitable for material classification under unstable illumination. In the following work, a multispectral 3D vision system was developed which can capture images of UV, red, green, blue and NIR simultaneously and then create denser 3D point clouds named multispectral 3D point clouds, in which image fusion algorithms and 3D morphological analysis can be conducted.
School/Discipline
University of South Australia. School of Engineering.
School of Engineering.
School of Engineering.
Dissertation Note
Thesis (PhD)--University of South Australia, 2018.
Provenance
Copyright 2018 Huajian Liu.
Description
1 ethesis (xv, 166 pages) :
illustrations (chiefly colour)
Includes bibliographical references (pages 154-166)
illustrations (chiefly colour)
Includes bibliographical references (pages 154-166)
Access Status
506 0#$fstar $2Unrestricted online access