Abstract

Weeds cost Australian farmers around $1.5 billion a year in weed control activities and a further $2.5 billion a year in lost agricultural production. Weed management requires a good understanding of weed inventories and distribution for effective management. Nowadays, cutting-edge research provides improved options for remote weed detection, facilitating broader adoption of these transformational technologies like airborne, drones, and satellites, to provide tools to improve weed management in complex environmental and agricultural systems. In this paper, we present our recent research work on applying two deep learning approaches to identify tiny weeds from airborne captured RGB images with the goal of determining feasible approaches for weeds managers. High accuracy and low false-positive have been achieved through convolutional network learning. To address the challenges remote sensing images had, such as low image resolution, high similarity, and a large volume of data, the deep learning-based approach shows superior performance to detect weeds in heterogeneous landscapes. Our findings will enhance remote sensing capabilities in the Australian weed community through knowledge and skills transfer and stimulate the development of applications to process.
Original languageEnglish
Title of host publicationImage and Video Technology
Subtitle of host publication10th Pacific-Rim Symposium, PSIVT 2022, Virtual Event, November 12–14, 2022, Proceedings
EditorsHan Wang, Paul Manoranjan, Kap Luk Chan, Guiju Ping, Wei Lin, Guobao Xiao, Xiaonan Wang, Haoge Jiang
Place of PublicationVirtual
PublisherSpringer
Pages159-171
Number of pages13
Volume13763
ISBN (Electronic)9783031264313
ISBN (Print)9783031264306
DOIs
Publication statusPublished - Apr 2023
Event10th Pacific-Rim Symposium on Image and Video Technology, PSIVT 2022 - Virtual
Duration: 25 Nov 202228 Nov 2022
https://link-springer-com.ezproxy.csu.edu.au/content/pdf/bfm:978-3-031-26431-3/1?pdf=chapter%20toc (Proceedings front matter)
http://www.cis-ram.org/psivt2022/workshops.html (Conference website)
https://link-springer-com.ezproxy.csu.edu.au/book/10.1007/978-3-031-26431-3 (Proceedings)

Publication series

NameLecture Notes in Computer Science
PublisherSpringer
Volume13763
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference10th Pacific-Rim Symposium on Image and Video Technology, PSIVT 2022
Period25/11/2228/11/22
OtherThe PSIVT is a premier level biennial series of symposia that aims to provide a forum for researchers and practitioners in the Pacific Rim and around the world who are involving in contributing to theoretical advances or practical implementations in image and video technology. The PSIVT has been held 9 times. It is a highly referenced conference that provides authors with useful feedbacks. Submissions are invited on significant, original, and previously unpublished research on all aspects of image and video technology. All papers will receive mindful and rigorous reviews.
Internet address

Fingerprint

Dive into the research topics of 'Remote tiny weeds detection'. Together they form a unique fingerprint.

Cite this