Abstract:

Objective As the global population continues to expand and climate change poses increasingly severe challenges, ensuring food security has become a paramount concern for the 21st century. In response to these pressing issues, there's a growing demand among plant cultivators and breeders for efficient methods to acquire plant phenotypic traits at high throughput, facilitating the establishment of mappings from phenotypes to genotypes. By integrating mobile phenotyping platforms with improved instance segmentation techniques, researchers have achieved a significant advancement in the automation and accuracy of phenotypic data extraction. Such developments are pivotal in accelerating the breeding process, improving crop resilience, and ultimately contributing to global food security efforts amidst the challenges posed by population growth and climate variability. Addressing the need for rapid extraction of leaf age and canopy width phenotypes in strawberry plants cultivated in controlled environments, this study introduces a novel high-throughput phenotyping extraction approach leveraging a mobile phenotyping platform and instance segmentation technology. Methods Data acquisition was conducted using a compact mobile phenotyping platform equipped with an array of sensors, including an RGB sensor, and edge control computers, capable of capturing overhead images of potted strawberry plants in greenhouses. Subsequent to previous work, targeted adjustments to the network structure were made to develop an enhanced Mask R-CNN (convolutional neural network) model for processing strawberry plant image data and rapidly extracting plant phenotypic information. The model initially employed a Split-Attention Networks (ResNeSt) backbone with a group attention module, replacing the original network to improve the precision and efficiency of image feature extraction. Moreover, during training, the model adopted the Mosaic method, suitable for instance segmentation data augmentation, to expand the dataset of strawberry images. Additionally, it optimized the original cross-entropy classification loss function with a binary cross-entropy loss function to achieve better detection accuracy of plants and leaves. Based on this, the improved Mask R-CNN description involves post-processing of training results. It utilized the positional relationship between leaf and plant masks to statistically count the number of leaves. Additionally, it employed segmentation masks and image calibration against true values to calculate the canopy width of the plant. Results and Discussions This research rigorously evaluated and compared the performance of an enhanced Mask R-CNN model, underpinned by the ResNeSt-101 backbone network. This model achieved a notable mask accuracy of 80.1% and a detection box accuracy of 89.6%. It was capable of performing high-throughput estimations for the age of strawberry leaves, demonstrating a high plant detection rate of 99.3% and a leaf count accuracy of 98.0%. This accuracy marked a significant improvement over the original Mask R-CNN model and meets the precise needs for phenotypic data extraction. The method exhibited considerable accuracy in measuring the canopy widths of strawberry plants, with errors less than 5% in about 98.1% of the cases, showcasing its effectiveness in phenotypic dimension assessment. Moreover, the model operated at a speed of 12.9 frames per second (FPS) on edge devices, striking a balance between accuracy and operational efficiency. This speed was sufficient for real-time applications and enabled rapid phenotypic data extraction even on devices with limited computational power. Compared to other instance segmentation models, this approach not only achieved higher accuracy in plant phenotypic data extraction but also maintained a good processing speed, making it well-suited for real-world agricultural applications. Its capability to extract plant phenotypic information quickly and accurately supported the advancement of smart and precision agriculture by facilitating the automation and intelligent management of crop monitoring and analysis. Conclusions This study has effectively deployed a mobile phenotyping platform coupled with instance segmentation techniques to process image data and extract diverse phenotypic indicators of strawberries. Notably, the method showcases remarkable robustness, in line with the fundamental tenets of smart and precision agriculture, which underscore the automation and intelligent oversight of agricultural processes. The seamless integration of mobile platforms and advanced image processing techniques not only enhances efficiency but also underscores a paradigm shift towards data-driven decision-making in agriculture. Ultimately, this research heralds a promising future for sustainable and optimized agricultural practices, where technological innovations play a pivotal role in meeting global food demands while minimizing environmental impact.

Key words: mobile phenotype platform, instance segmentation, strawberry, leaf age, plant crown width, plant phenotype