Click on “Download PDF” for the PDF version or on the title for the HTML version.
If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.
Active Learning for Real-Time Flower Counting with a Ground Mobile Robot
Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org
Citation: 2024 ASABE Annual International Meeting 2400607.(doi:10.13031/aim.202400607)
Authors: Daniel J Petti, Changying Li
Keywords: High-Throughput Phenotyping, Computer Vision, Edge Computing, Ground Mobile Robot
Abstract. Modern computer vision has made great strides in object recognition and counting, which have slowly filtered into the agricultural domain. Cotton flower counting is a good example of this. Though season-long flower counts have value to plant breeders, a manual data collection process is too laborious to be practical in most cases. In recent years, several fully automated flower counting approaches have been proposed. However, such approaches are typically designed to run offline and require a significant amount of computation. Furthermore, little thought has gone into developing convenient interfaces and integrations so that a layperson can use such systems without extensive training. The goal of this study is twofold: First, we use self-supervised representations to build a strong, black-box active learning framework. We then adopt this framework in order to build a lightweight flower tracking model that is deployable on a ground robot and can operate in real-time. Second, by using camera and GPS data from the robot, we extract flower locations automatically. We show that our approach can achieve <10% MAPE in flower counts while running in real-time on an Nvidia Jetson Xavier AGX. Overall, we believe that our highly integrated, automated, and simplified flower counting solution makes significant strides towards a practical commercial cotton phenotyping platform.
(Download PDF) (Export to EndNotes)
|