Unsupervised domain adaptation for highlight detection and removal in agricultural robot vision system.

Saved in:

Bibliographic Details
Title:	Unsupervised domain adaptation for highlight detection and removal in agricultural robot vision system.
Authors:	Laide Guan¹ guanlaide@gmail.com, Bole Li¹
Source:	Journal of Biotech Research. 2024, Vol. 19, p355-364. 10p.
Subjects:	Robot vision, Ubuntu (Operating system), Agricultural robots, Feature extraction, Agriculture
Abstract:	Assessment of the cotton boll health is essential for field oversight and maturity rating. Typical methods for agricultural image acquiring are unsuitable for cotton boll images due to the possibility of distribution mismatch that resulted from environmental variables such as duration, climate, farming activities, and regions. Adapting a domain can solve this problem. This study applied a domain-adversarial neural network-driven unsupervised domain adaptation (DANN-UDA) approach to gather the cotton bolls datasets, which involved multiple steps of target label inference and dense inherent ConvNet-based feature extraction. The proposed approach was executed in the agricultural robot with Ubuntu and Robot Operating System (ROS) environment and verified using an agricultural robot captured cotton boll image. The efficiency of proposed DANN-UDA method in cotton boll identification was evaluated and demonstrated better results. The performance of the cotton boll stage detection of proposed DANN-UDA was compared using Visual Geometry Group (VGG) and You Only Look Once version 5 (YOLOv5) networks in agricultural robot vision system. The results demonstrated that the proposed approach obtained the best identification outcomes in a variety of scenarios. Additionally, the proposed model could serve as a helpful substitute for human observation and conventional categorization techniques. [ABSTRACT FROM AUTHOR]
	Copyright of Journal of Biotech Research is the property of Bio Tech System and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	Assessment of the cotton boll health is essential for field oversight and maturity rating. Typical methods for agricultural image acquiring are unsuitable for cotton boll images due to the possibility of distribution mismatch that resulted from environmental variables such as duration, climate, farming activities, and regions. Adapting a domain can solve this problem. This study applied a domain-adversarial neural network-driven unsupervised domain adaptation (DANN-UDA) approach to gather the cotton bolls datasets, which involved multiple steps of target label inference and dense inherent ConvNet-based feature extraction. The proposed approach was executed in the agricultural robot with Ubuntu and Robot Operating System (ROS) environment and verified using an agricultural robot captured cotton boll image. The efficiency of proposed DANN-UDA method in cotton boll identification was evaluated and demonstrated better results. The performance of the cotton boll stage detection of proposed DANN-UDA was compared using Visual Geometry Group (VGG) and You Only Look Once version 5 (YOLOv5) networks in agricultural robot vision system. The results demonstrated that the proposed approach obtained the best identification outcomes in a variety of scenarios. Additionally, the proposed model could serve as a helpful substitute for human observation and conventional categorization techniques. [ABSTRACT FROM AUTHOR]
ISSN:	19443285