Revolutionizing strawberry seedling management with automated detection system

This method, which utilizes advanced deep learning techniques, aims to optimize seedling management in glass greenhouses, ensuring uniform growth and maximizing yields.

Strawberries are a high-value crop, and maintaining virus-free mother plants is essential to ensure genetic stability and prevent virus transmission. In large-scale greenhouse environments, virus-free strawberry mother plants are grown in rectangular nutrient pots, which help optimize plant growth conditions. However, over time, seedlings may go missing due to environmental factors, leading to lower yields and inconsistent quality. Timely detection of missing seedlings is critical to prevent these losses. Traditionally, manual inspection has been used, but it is labor-intensive and prone to error. The need for a precise, automated system to detect missing seedlings and monitor seedling quality has led to the development of SSP-MambaNet, a deep learning-based solution designed to address these challenges.

A study (DOI: 10.1016/j.plaphe.2025.100043) published in Plant Phenomics on 9 April 2025 by Jifeng Ning’s team, Northwest A&F University, enhances operational efficiency and supports high-quality strawberry production in large-scale, automated greenhouse environments.

The research team conducted a series of experiments to evaluate the performance of SSP-MambaNet in detecting strawberry seedlings and nutrient pots. Using an Intel Xeon E5-2620 CPU and an NVIDIA GeForce RTX 2080Ti GPU, the team implemented the system on Ubuntu 16.04, with Visual Studio 2017, Python 3.8, and PyTorch 1.8 for development. The team compared SSP-MambaNet against several established models, including SSD, Faster R-CNN, YOLOv5, YOLOv7, and others, using a standardized input image resolution of 640×640. Training involved a batch size of 50, a learning rate that transitioned from 0.01 to 0.1, and specific parameters to balance model performance. The results revealed that SSP-MambaNet outperformed other models, achieving significant improvements in mean average precision (mAP) across multiple benchmarks—5.1% over YOLOv5, 4.4% over YOLOX, and 25.1% over RT-DETR-X. Additionally, it achieved the highest F1 score of 90.4 and a frame rate of 66.7 frames per second, highlighting its superior speed and accuracy. The team further conducted ablation experiments, showing that modules such as SPDFFA and CVSSB improved model precision and recall, particularly under conditions with dense occlusions and complex lighting. For instance, SPDFFA enhanced feature retention in low-light and occluded areas, while CVSSB-E boosted understanding of complex scenes. The introduction of MPDIoU further refined localization accuracy. Moreover, the model's detection of missing seedlings was particularly impressive, with a 94.29% accuracy rate, demonstrating the system's potential for timely intervention in strawberry cultivation. Overall, SSP-MambaNet proved to be a highly effective and efficient tool for managing strawberry seedling deficiencies, offering significant improvements over traditional methods.

SSP-MambaNet provides a cutting-edge solution to the problem of missing strawberry seedlings in greenhouse environments. By combining deep learning with advanced feature fusion and localization techniques, this system achieves impressive accuracy in identifying missing seedlings and counting nutrient pots with missing plants. The method not only improves the management of strawberry seedlings but also holds potential for broader agricultural applications. As greenhouse farming continues to evolve, automated systems like SSP-MambaNet will play an essential role in enhancing production efficiency and ensuring high-quality crop yields.

###

References

DOI

10.1016/j.plaphe.2025.100043

Original URL

https://doi.org/10.1016/j.plaphe.2025.100043

Funding information

This work was supported by the National Natural Science Foundation of China (62476227).

About Plant Phenomics

Plant Phenomics is dedicated to publishing novel research that will advance all aspects of plant phenotyping from the cell to the plant population levels using innovative combinations of sensor systems and data analytics. Plant Phenomics aims also to connect phenomics to other science domains, such as genomics, genetics, physiology, molecular biology, bioinformatics, statistics, mathematics, and computer sciences. Plant Phenomics should thus contribute to advance plant sciences and agriculture/forestry/horticulture by addressing key scientific challenges in the area of plant phenomics.