Narrow and broad weed classification using scale invariant features transform

Kamarul Hawari Ghazali, Mohd. Marzuki Mustafa, Aini Hussain, Siti Nur Hafizah MohdZaid

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Scale Invariant Feature Extraction (SIFT) has been proven to be the most robust technique for object detection. The features are invariant to image scale and rotation and are shown to provide good matching across a substantial range of affine distortion, addition of noise, change in 3D viewpoint and change in illumination. Implementing SIFT onto an image will produce a large values of data (key points) that represent the magnitude of the local gradient with respect to its orientation. We introduce a new technique using SIFT to classify our target object, narrow and broad weed for weeding strategy in oil palm plantation. The SIFT for classification technique (SIFT-C) was derived at magnitude and orientation histogram of local descriptor and its feature vector has been extracted and used as an input for classifier design. Linear Discriminant System was used as a classification tools to determine the linear equation of classifier system. The classification result shows that narrow and broad weed were successfully identified with an average of correct classification rate of above 97%.

Original languageEnglish
Title of host publicationIET Conference Publications
Pages239-246
Number of pages8
Edition529 CP
DOIs
Publication statusPublished - 2007
EventChina-Ireland International Conference on Information and Communications Technologies, CIICT 2007 - Dublin
Duration: 28 Aug 200729 Aug 2007

Other

OtherChina-Ireland International Conference on Information and Communications Technologies, CIICT 2007
CityDublin
Period28/8/0729/8/07

Fingerprint

Feature extraction
Classifiers
Palm oil
Linear equations
Lighting

Keywords

  • Invariant
  • Linear discriminant system
  • SIFT
  • Weed

ASJC Scopus subject areas

  • Electrical and Electronic Engineering

Cite this

Ghazali, K. H., Mustafa, M. M., Hussain, A., & MohdZaid, S. N. H. (2007). Narrow and broad weed classification using scale invariant features transform. In IET Conference Publications (529 CP ed., pp. 239-246) https://doi.org/10.1049/cp:20070705

Narrow and broad weed classification using scale invariant features transform. / Ghazali, Kamarul Hawari; Mustafa, Mohd. Marzuki; Hussain, Aini; MohdZaid, Siti Nur Hafizah.

IET Conference Publications. 529 CP. ed. 2007. p. 239-246.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Ghazali, KH, Mustafa, MM, Hussain, A & MohdZaid, SNH 2007, Narrow and broad weed classification using scale invariant features transform. in IET Conference Publications. 529 CP edn, pp. 239-246, China-Ireland International Conference on Information and Communications Technologies, CIICT 2007, Dublin, 28/8/07. https://doi.org/10.1049/cp:20070705
Ghazali, Kamarul Hawari ; Mustafa, Mohd. Marzuki ; Hussain, Aini ; MohdZaid, Siti Nur Hafizah. / Narrow and broad weed classification using scale invariant features transform. IET Conference Publications. 529 CP. ed. 2007. pp. 239-246
@inproceedings{80f988db9ab04341bdf5472cd8a5608b,
title = "Narrow and broad weed classification using scale invariant features transform",
abstract = "Scale Invariant Feature Extraction (SIFT) has been proven to be the most robust technique for object detection. The features are invariant to image scale and rotation and are shown to provide good matching across a substantial range of affine distortion, addition of noise, change in 3D viewpoint and change in illumination. Implementing SIFT onto an image will produce a large values of data (key points) that represent the magnitude of the local gradient with respect to its orientation. We introduce a new technique using SIFT to classify our target object, narrow and broad weed for weeding strategy in oil palm plantation. The SIFT for classification technique (SIFT-C) was derived at magnitude and orientation histogram of local descriptor and its feature vector has been extracted and used as an input for classifier design. Linear Discriminant System was used as a classification tools to determine the linear equation of classifier system. The classification result shows that narrow and broad weed were successfully identified with an average of correct classification rate of above 97{\%}.",
keywords = "Invariant, Linear discriminant system, SIFT, Weed",
author = "Ghazali, {Kamarul Hawari} and Mustafa, {Mohd. Marzuki} and Aini Hussain and MohdZaid, {Siti Nur Hafizah}",
year = "2007",
doi = "10.1049/cp:20070705",
language = "English",
isbn = "9780863418273",
pages = "239--246",
booktitle = "IET Conference Publications",
edition = "529 CP",

}

TY - GEN

T1 - Narrow and broad weed classification using scale invariant features transform

AU - Ghazali, Kamarul Hawari

AU - Mustafa, Mohd. Marzuki

AU - Hussain, Aini

AU - MohdZaid, Siti Nur Hafizah

PY - 2007

Y1 - 2007

N2 - Scale Invariant Feature Extraction (SIFT) has been proven to be the most robust technique for object detection. The features are invariant to image scale and rotation and are shown to provide good matching across a substantial range of affine distortion, addition of noise, change in 3D viewpoint and change in illumination. Implementing SIFT onto an image will produce a large values of data (key points) that represent the magnitude of the local gradient with respect to its orientation. We introduce a new technique using SIFT to classify our target object, narrow and broad weed for weeding strategy in oil palm plantation. The SIFT for classification technique (SIFT-C) was derived at magnitude and orientation histogram of local descriptor and its feature vector has been extracted and used as an input for classifier design. Linear Discriminant System was used as a classification tools to determine the linear equation of classifier system. The classification result shows that narrow and broad weed were successfully identified with an average of correct classification rate of above 97%.

AB - Scale Invariant Feature Extraction (SIFT) has been proven to be the most robust technique for object detection. The features are invariant to image scale and rotation and are shown to provide good matching across a substantial range of affine distortion, addition of noise, change in 3D viewpoint and change in illumination. Implementing SIFT onto an image will produce a large values of data (key points) that represent the magnitude of the local gradient with respect to its orientation. We introduce a new technique using SIFT to classify our target object, narrow and broad weed for weeding strategy in oil palm plantation. The SIFT for classification technique (SIFT-C) was derived at magnitude and orientation histogram of local descriptor and its feature vector has been extracted and used as an input for classifier design. Linear Discriminant System was used as a classification tools to determine the linear equation of classifier system. The classification result shows that narrow and broad weed were successfully identified with an average of correct classification rate of above 97%.

KW - Invariant

KW - Linear discriminant system

KW - SIFT

KW - Weed

UR - http://www.scopus.com/inward/record.url?scp=67650115230&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=67650115230&partnerID=8YFLogxK

U2 - 10.1049/cp:20070705

DO - 10.1049/cp:20070705

M3 - Conference contribution

SN - 9780863418273

SP - 239

EP - 246

BT - IET Conference Publications

ER -