Automatic root measurement: a lightweight method for measuring pea root length.

Abstract

Background With the intensification of global climate change, extreme weather events have become increasingly frequent, severely impacting the growth cycles and yield stability of crops. Against this backdrop, cultivating new crop varieties with high stress resistance has become a core task for achieving sustainable agriculture and ensuring food security. Root length, as a critical phenotypic trait that reflects a plant's ability to absorb water and nutrients, is closely related to the crop's capacity to withstand adversities, such as drought, high temperatures and salinisation. However, root length measurement technology remains a significant bottleneck in plant science research. Traditional manual methods are inefficient and prone to human-induced variability (e.g. subjective standard discrepancies, operational errors, and potential contamination or damage to seeds). Meanwhile, existing automated measurement models face challenges in large-scale practical applications due to their high deployment costs. Results This study developed a seed germination image acquisition system and constructed a pea root dataset. Based on the YOLOv8-Seg-n instance segmentation model, a lightweight automatic root measurement (ARM) model was then developed using feature distillation, structured pruning techniques, and a series of post-processing procedures for root length calculation. Experimental results demonstrated that the ARM model had only 1.81 M parameters, with 8.3 GFLOPs and a weight file size of 4.2 MB, and achieved 70.4 FPS. It realised outstanding performance with mAP@0.5 and AP root scores of 90.3% and 81.2%, respectively, showing a high consistency with manual measurement results (R² = 0.993). Compared to existing models, the ARM model significantly reduces parameter scale and computational complexity, making it more accommodating to device performance and computational requirements while also decreasing the workload associated with root sample processing. Furthermore, the application of the ARM model in a 72-hour full time-series analysis of pea root length under drought conditions validated its potential for practical use in real-world scenarios. Conclusions The ARM model offers an efficient and cost-effective technological solution for high-throughput root length measurement in peas. It achieves a favorable balance between accuracy, speed, and computational resource requirements, demonstrating broad application potential in agricultural production and breeding research. The model offers critical technical support for ensuring food security and enhancing crop stress resistance.

Plant phenotyping relevance

エンドツーエンドの画像取得・セグメンテーション・根長算出モデルを開発し、手動測定との整合性および実利用を検証しており、植物表現型測定法が中心である。

Code and data availability

The supplied blocks describe a pea root dataset and a lightweight ARM model based on YOLOv8-Seg, but contain no data or code availability statement, no public repository URL for the authors' dataset, images, code, or trained model. The only GitHub links (ultralytics/yolov5, ultralytics/ultralytics) are third-party base

No evidence-backed public reproduction asset is currently recorded.