YOLOv9-GSSA model for efficient soybean seedlings and weeds detection

To monitor soybean seedlings growth in real time, an effective method for accurately identifying seedlings and removing weeds is essential. Challenges include the small size and morphological similarity of seedlings and weeds, complicating conventional detection methods. To tackle these issues, we p...

Full description

Saved in:
Bibliographic Details
Main Authors: Baihe Liang, Liangchen Hu, Guangxing Liu, Peng Hu, Shaosheng Xu, Biao Jie
Format: Article
Language:English
Published: Elsevier 2025-12-01
Series:Smart Agricultural Technology
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2772375525003661
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849702524465971200
author Baihe Liang
Liangchen Hu
Guangxing Liu
Peng Hu
Shaosheng Xu
Biao Jie
author_facet Baihe Liang
Liangchen Hu
Guangxing Liu
Peng Hu
Shaosheng Xu
Biao Jie
author_sort Baihe Liang
collection DOAJ
description To monitor soybean seedlings growth in real time, an effective method for accurately identifying seedlings and removing weeds is essential. Challenges include the small size and morphological similarity of seedlings and weeds, complicating conventional detection methods. To tackle these issues, we propose a real-time detection algorithm called YOLOv9-GSSA. The improved Mosaic-Dense algorithm increases object density at the model's input layer, enhancing its ability to capture detailed features. Additionally, the GSSA neck optimization module, combining GSConv and Gated Self-Attention, supports key information extraction and multi-scale feature interaction. The Swin-GSSA prediction head further utilizes spatial positional information, improving small object detection and handling overlapping occlusion. Experimental results show our model achieves a mAP of 47.5% with a detection speed of 23.42 ms per image, suitable for real-time monitoring. The enhanced model significantly improves the detection of soybean seedlings and weeds, making it a valuable tool for managing farmland effectively. This ultimately aids in precise yield estimation and decision-making in precision agriculture.
format Article
id doaj-art-261bfccd7bf54db79fcb3e7e6534247e
institution DOAJ
issn 2772-3755
language English
publishDate 2025-12-01
publisher Elsevier
record_format Article
series Smart Agricultural Technology
spelling doaj-art-261bfccd7bf54db79fcb3e7e6534247e2025-08-20T03:17:38ZengElsevierSmart Agricultural Technology2772-37552025-12-011210113410.1016/j.atech.2025.101134YOLOv9-GSSA model for efficient soybean seedlings and weeds detectionBaihe Liang0Liangchen Hu1Guangxing Liu2Peng Hu3Shaosheng Xu4Biao Jie5School of Computer and Information, Anhui Normal University, Wuhu 241002, ChinaSchool of Computer and Information, Anhui Normal University, Wuhu 241002, China; Anhui Provincial Key Laboratory of Industrial Intelligence Data Security, Wuhu 241002, China; Corresponding author at: School of Computer and Information, Anhui Normal University, Wuhu 241002, China.School of Computer and Information, Anhui Normal University, Wuhu 241002, ChinaSchool of Computer and Information, Anhui Normal University, Wuhu 241002, China; Anhui Provincial Key Laboratory of Industrial Intelligence Data Security, Wuhu 241002, ChinaSchool of Computer and Information, Anhui Normal University, Wuhu 241002, China; Anhui Provincial Key Laboratory of Industrial Intelligence Data Security, Wuhu 241002, ChinaSchool of Computer and Information, Anhui Normal University, Wuhu 241002, China; Anhui Provincial Key Laboratory of Industrial Intelligence Data Security, Wuhu 241002, ChinaTo monitor soybean seedlings growth in real time, an effective method for accurately identifying seedlings and removing weeds is essential. Challenges include the small size and morphological similarity of seedlings and weeds, complicating conventional detection methods. To tackle these issues, we propose a real-time detection algorithm called YOLOv9-GSSA. The improved Mosaic-Dense algorithm increases object density at the model's input layer, enhancing its ability to capture detailed features. Additionally, the GSSA neck optimization module, combining GSConv and Gated Self-Attention, supports key information extraction and multi-scale feature interaction. The Swin-GSSA prediction head further utilizes spatial positional information, improving small object detection and handling overlapping occlusion. Experimental results show our model achieves a mAP of 47.5% with a detection speed of 23.42 ms per image, suitable for real-time monitoring. The enhanced model significantly improves the detection of soybean seedlings and weeds, making it a valuable tool for managing farmland effectively. This ultimately aids in precise yield estimation and decision-making in precision agriculture.http://www.sciencedirect.com/science/article/pii/S2772375525003661GSConvMosaicObject detectionSelf-attentionYOLO
spellingShingle Baihe Liang
Liangchen Hu
Guangxing Liu
Peng Hu
Shaosheng Xu
Biao Jie
YOLOv9-GSSA model for efficient soybean seedlings and weeds detection
Smart Agricultural Technology
GSConv
Mosaic
Object detection
Self-attention
YOLO
title YOLOv9-GSSA model for efficient soybean seedlings and weeds detection
title_full YOLOv9-GSSA model for efficient soybean seedlings and weeds detection
title_fullStr YOLOv9-GSSA model for efficient soybean seedlings and weeds detection
title_full_unstemmed YOLOv9-GSSA model for efficient soybean seedlings and weeds detection
title_short YOLOv9-GSSA model for efficient soybean seedlings and weeds detection
title_sort yolov9 gssa model for efficient soybean seedlings and weeds detection
topic GSConv
Mosaic
Object detection
Self-attention
YOLO
url http://www.sciencedirect.com/science/article/pii/S2772375525003661
work_keys_str_mv AT baiheliang yolov9gssamodelforefficientsoybeanseedlingsandweedsdetection
AT liangchenhu yolov9gssamodelforefficientsoybeanseedlingsandweedsdetection
AT guangxingliu yolov9gssamodelforefficientsoybeanseedlingsandweedsdetection
AT penghu yolov9gssamodelforefficientsoybeanseedlingsandweedsdetection
AT shaoshengxu yolov9gssamodelforefficientsoybeanseedlingsandweedsdetection
AT biaojie yolov9gssamodelforefficientsoybeanseedlingsandweedsdetection