The advent of automated image analysis has ushered in a new era in plant research, particularly in the high-throughput phenotyping of legume traits. Legumes, vital contributors to global agriculture and food security
, demand efficient and accurate phenotyping methodologies for crop improvement programs. Traditional phenotyping methods
- reliant on manual measurements, pose challenges in terms of scalability and precision. In response, automated image analysis has emerged as a transformative solution, leveraging advancements in machine learning and computer vision. Use of digital imaging technology is prevalent in the evaluation of legume seed traits and has proven to be useful in high-throughput phenotyping
(Margapuri et al., 2021; Shariff et al., 2010). ImageJ, Cell Profiler, Win SEEDLE, Smart Grain and P-TRAP are open-source software applications that utilize digital imaging technologies to rapidly calculate the 2D characteristics of seeds. However, digital imaging technology has challenges in capturing 3D characteristics, including volume, surface area thickness. Vision technologies and 3D reconstruction technology are utilized across diverse engineering disciplines.
Chen et al., (2020) examined the increased 3D perception of the central stock of orchard bananas using adaptive multisession technology
(Chen et al., 2020). Firatligil-Durmuş et al., (2010) identified the round or elongated fruits on plants in their natural surroundings and directed harvesting robots to automatically collect them using a 3D fruit detection algorithm that relies on color, depth and shape. Research in agriculture is actively exploring the use of three-dimensional (3D) technologies for measurement purposes
(Yang et al., 2020). The technique can be utilized for quantifying leaf area, leaf angle, stems and shoots, fruit and seeds
(Miao et al., 2021). Due to the large sample sizes employed in high-throughput phenotyping analysis, it is crucial to explore faster and more automated methods for processing data in batches
(Xu et al., 2018). Soybeans, peas, black beans, red beans and mung beans are common legume seeds that hold significant dietary value Globally. The process of high-throughput legume seed phenotyping is highly important as it enables a more convenient assessment of both the yield and quality of legume seeds. Legume seeds exhibit a diverse range of forms and sizes. The typical shapes of these objects are roughly spherical or ellipsoidal
(Cervantes and Martín Gómez, 2019). Soybeans and peas exemplify legume seeds characterized by spherical and ellipsoidal shapes. Notable examples of ellipsoidal legume seeds include black beans, red beans and mung beans.
(Xu et al., 2018). Spherical or elliptical seeds possess symmetry, which can be utilized to expedite batch 3D modeling
(Yang et al., 2021).
This article examined the growing body of research focused on employing automated image processing for legume phenotyping. The critical need for accurate phenotypic characterization in legumes is emphasized given their multifaceted roles, encompassing improvements in nutritional content and heightened resilience amid shifting environmental conditions. Placing our study within the wider context of plant phenotyping, the introduction draws attention to the unique challenges faced by researchers investigating legumes and underscores the necessity for innovative methodologies to address these challenges.
Plant phenotyping provides the basis for comprehending how plants react to environmental pressures and genetic differences, which in turn helps in enhancing crops and ensuring agricultural sustainability. Chlorophyll fluorescence imaging is a valuable technique for evaluating the physiological condition of plants without causing damage. It provides valuable information on the efficiency of photosynthesis and the ability of plants to withstand stress. However, manually examining fluorescence images is a laborious process that is sensitive to human error. Consequently, there is a growing use of sophisticated image processing methods to automate the analysis procedure and get quantitative data from fluorescence pictures. This paper presents a complete method for analysing chlorophyll fluorescence images and measuring important fluorescence metrics, such as Fv/Fm and NPQ, in order to improve our knowledge of how plants respond to environmental stimuli.
Legumes, encompassing a diverse group of plants such as soybeans, chickpeas and lentils, play a pivotal role in global agriculture due to their nutritional value, soil-enriching properties symbiotic nitrogen fixation. Traditional phenotyping methods for legume traits have long relied on manual measurements, but limitations in scalability and precision have driven a quest for innovative approaches. Historically, legume phenotyping faced challenges in capturing the intricacies of morphological traits critical for crop improvement. Conventional methods, often labor-intensive and time-consuming, struggled to keep pace with the demands for high-throughput analysis
(Yang et al., 2020). Recognizing these limitations, recent literature showcases a surge in studies employing automated image analysis techniques. Machine learning algorithms, including support vector machines (SVM) and random forests
, demonstrate efficacy in accurately classifying and quantifying morphological traits such as leaf size, shape and canopy architecture. The literature highlights their ability to process large datasets rapidly, enabling efficient phenotypic characterization
(Elbasi et al., 2023; Cho, 2024;
Kim and AlZubi, 2024;
Min et al., 2024; Porwal et al., 2024; Wasik and Pattinson, 2024;
Maltare, 2023).
Deep learning approaches, particularly convolutional neural networks (CNNs), have gained prominence in image-based phenotyping. Their capacity to learn hierarchical features and recognize complex patterns makes them invaluable for legume trait assessment. Studies showcase successful applications of CNNs in identifying nuanced traits, from nodulation patterns to subtle changes in leaf color indicative of stress conditions
(Yu et al., 2023). The adaptability of deep learning to diverse legume species emphasizes its potential as a universal tool for comprehensive phenotyping. In parallel, traditional computer vision methods, though eclipsed by machine learning and deep learning, continue to find relevance in certain applications. Studies highlight their utility in addressing specific challenges, such as the precise quantification of root traits through advanced image segmentation techniques
(Wang and Su, 2022). The literature review elucidates how these methods contribute to a holistic understanding of legume phenotypes
-complementing the strengths of machine learning and deep learning approaches.
As the literature unfolds, it becomes evident that automated image analysis extends beyond morphological trait assessment. Physiological parameters, including chlorophyll content, stomatal conductance water use efficiency, emerge as focal points in legume phenotyping. The integration of these parameters into automated systems facilitates an understanding of plant health, stress responses and overall crop performance
(Bertolino et al., 2019). These insights are crucial for breeding programs aiming to develop legume varieties that are resilient to changing environmental conditions. Moreover, the literature underlines the significance of automated image analysis in stress and disease detection within legume crops
(Warman et al., 2021). Early identification of stressors facilitated by sophisticated algorithms and image-based diagnostics, equips researchers and farmers with timely information for targeted interventions
(Holzinger et al., 2023). The potential for automated systems to contribute to sustainable agriculture by minimizing the reliance on chemical interventions aligns with global efforts towards environmentally conscious crop management
(Balaska et al., 2023).
