Full-time sequence assessment of okra seedling vigor under salt stress based on leaf area and leaf growth rate estimation using the YOLOv11-HSECal instance segmentation model.

Abstract

Introduction With the growing severity of global salinization, assessing plant growth vitality under salt stress has become a critical aspect in agricultural research. Methods In this paper, a method for calculating the leaf area and leaf growth rate of okra based on the YOLOv11-HSECal model is proposed, which is used to evaluate the activity of okra at the seedling stage. A high-throughput, Full-Time Sequence Crop Germination Vigor Monitoring System was developed to automatically capture image data from seed germination to seedling growth stage, while maintaining stable temperature and lighting conditions. To address the limitations of the traditional YOLOv11-seg model, the YOLOv11-HSECal model was optimized by incorporating the HGNetv2 backbone, Slim-Neck feature fusion, and EMAttention mechanisms. Results These improvements led to a 1.1% increase in mAP50, a 0.6% reduction in FLOPs, and a 14.1% decrease in model parameters. Additionally, Merge and Cal modules were integrated for calculating the leaf area and growth rate of okra seedlings. Finally, through salt stress experiments, we assessed the effects of varying NaCl concentrations (CK, 10 mmol/L, 20 mmol/L, 30 mmol/L, 40 mmol/L, 50 mmol/L, and 60 mmol/L) on the leaf area and growth rate of okra seedlings, verifying the inhibitory effects of salt stress on seedling vitality. Discussion The results demonstrate that the YOLOv11-HSECal model efficiently and accurately evaluates okra seedling growth vitality under salt stress in a full-time monitoring manner, offering significant potential for broader applications. This work provides a novel solution for full-time plant growth monitoring and vitality assessment in smart agriculture and offers valuable insights into the impact of salt stress on crop growth.

Plant phenotyping relevance

YOLOベースの画像解析モデルと自動撮像システムを開発し、オクラ幼苗の葉面積・葉成長率を連続推定することが研究の中心であるため。

Code and data availability

The supplied blocks describe an okra seedling image dataset (1,723 valid time-series images, augmented to 2,707) and a YOLOv11-HSECal model, but contain no data availability statement, repository deposit, or authors' public URL for the dataset, images, code, or trained model. No paper-specific public asset is available

No evidence-backed public reproduction asset is currently recorded.