A Soybean Pod Accuracy Detection and Counting Model Based on Improved YOLOv8

Xiaofei Jia, Zhenlu Hua, Hongtao Shi, Dan Zhu, Zhongzhi Han, Guangxia Wu and Limiao Deng ()
Additional contact information
Xiaofei Jia: School of Science and Information, Qingdao Agricultural University, Qingdao 266109, China
Zhenlu Hua: School of Science and Information, Qingdao Agricultural University, Qingdao 266109, China
Hongtao Shi: School of Science and Information, Qingdao Agricultural University, Qingdao 266109, China
Dan Zhu: Key Lab of Plant Biotechnology in Universities of Shandong Province, College of Life Science, Qingdao Agricultural University, Qingdao 266109, China
Zhongzhi Han: School of Science and Information, Qingdao Agricultural University, Qingdao 266109, China
Guangxia Wu: School of Agronomy, Qingdao Agricultural University, Qingdao 266109, China
Limiao Deng: School of Science and Information, Qingdao Agricultural University, Qingdao 266109, China

Agriculture, 2025, vol. 15, issue 6, 1-27

Abstract: The number of soybean pods is a key determinant of soybean yield, making accurate detection and counting essential for yield estimation, cultivation management, and variety selection. Traditional manual counting methods are labor-intensive and time-consuming, and while object detection networks are widely applied in agricultural tasks, the dense distribution and overlapping occlusion of soybean pods present significant challenges. This study developed a soybean pod detection model, YOLOv8n-POD, based on the YOLOv8n network, incorporating key innovations to address these issues. A Dense Block Backbone (DBB) enhances the model’s adaptability to the morphological diversity of soybean pods, while the Separated and Enhancement Attention Module (SEAM) in the neck section improves the representation of pod-related features in feature maps. Additionally, a Dynamic Head increases the flexibility in detecting pods of varying scales. The model achieved an average precision (AP) of 83.1%, surpassing mainstream object detection methodologies with a 5.3% improvement over YOLOv8. Tests on three public datasets further demonstrated its generalizability to other crops. The proposed YOLOv8n-POD model provides robust support for accurate detection and localization of soybean pods, essential for yield estimation and breeding strategies, and its significant theoretical and practical implications extend its applicability to other crop types, advancing agricultural automation and precision farming.

Keywords: deep learning; image processing; object detection; YOLOv8; pod detection (search for similar items in EconPapers)
JEL-codes: Q1 Q10 Q11 Q12 Q13 Q14 Q15 Q16 Q17 Q18 (search for similar items in EconPapers)
Date: 2025
References: View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/2077-0472/15/6/617/pdf (application/pdf)
https://www.mdpi.com/2077-0472/15/6/617/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jagris:v:15:y:2025:i:6:p:617-:d:1612083

Access Statistics for this article

Agriculture is currently edited by Ms. Leda Xuan

More articles in Agriculture from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().