Canola has a prominent floral signature and requires careful consideration when selecting spectral indices for yield estimation. This study evaluated several spectral indices derived from high-resolution RGB images. A small plot (2.75m x 6m) experiment was conducted at Kernen Research Farm, Saskatoon, where canola was grown under varying row spacings and seeding rates (192 plots). The canopy reflectance was imaged during the flowering period and seed yield was obtained at physiological maturity. Indices were evaluated for accuracy in quantifying canola flowers in high-resolution RGB imagery with within-canopy shadow pixels. Digitalized flower number from the peak flowering date was used to test and validate a non-linear three-parameter asymptotic regression model to simulate canola seed yield. 70 % of the data was used to develop the model, and 30 % was used to validate the model. Model performance was tested with Pseudo-R2, r, MAE, and RMSE. HrFI (High-resolution Flowering Index) and MYI (Modified Yellowness Index) were able to accurately identify flowering pixels with the least amount of error pixel. The yield simulation model resulted in a pseudo-R2 value of 0.11 for the tested model and, a correlation of 0.91 for validation with RMSE and MAE of 343.1 and 265.3, respectively. Our results indicate that the HrFI index is a better indicator of yield potential compared to NDYI as the metric is well capable of handling within canopy shadows. Further studies are necessary to evaluate the performance of HrFI for medium resolution-UAV and satellite imagery.

Mung bean [Vigna radiata (L.) Wilczek] is a drought-tolerant, short-duration crop, and a rich source of protein and other valuable minerals, vitamins, and antioxidants. The main objectives of this research were (1) to study the root traits related with the phenotypic and genetic diversity of 375 mung bean genotypes of the Iowa (IA) diversity panel and (2) to conduct genome-wide association studies of root-related traits using the Automated Root Image Analysis (ARIA) software. We collected over 9,000 digital images at three-time points (days 12, 15, and 18 after germination). A broad sense heritability for days 15 (0.22–0.73) and 18 (0.23–0.87) was higher than that for day 12 (0.24–0.51). We also reported root ideotype classification, i.e., PI425425 (India), PI425045 (Philippines), PI425551 (Korea), PI264686 (Philippines), and PI425085 (Sri Lanka) that emerged as the top five in the topsoil foraging category, while PI425594 (unknown origin), PI425599 (Thailand), PI425610 (Afghanistan), PI425485 (India), and AVMU0201 (Taiwan) were top five in the drought-tolerant and nutrient uptake “steep, cheap, and deep” ideotype. We identified promising genotypes that can help diversify the gene pool of mung bean breeding stocks and will be useful for further field testing. Using association studies, we identified markers showing significant associations with the lateral root angle (LRA) on chromosomes 2, 6, 7, and 11, length distribution (LED) on chromosome 8, and total root length-growth rate (TRL_GR), volume (VOL), and total dry weight (TDW) on chromosomes 3 and 5. We discussed genes that are potential candidates from these regions. We reported beta-galactosidase 3 associated with the LRA, which has previously been implicated in the adventitious root development via transcriptomic studies in mung bean. Results from this work on the phenotypic characterization, root-based ideotype categories, and significant molecular markers associated with important traits will be useful for the marker-assisted selection and mung bean improvement through breeding.

For field workers around the world, wheat trials are often synonymous with wheat heads counting: a tedious but important task to measure this important yield component. Deep Learning has been a promising solution to automate the acquisition of wheat head density from a high-throughput phenotyping system, but it has been shown to be sensitive to changing acquisition conditions, also known as “domain change.” In response, an international collaboration built the “Global Wheat Head Dataset” in 2020 and 2021, a collection of 6515 images acquired during 47 different acquisition sessions in 12 countries. In addition to these datasets, two data competitions were held in 2020 (Kaggle, over 2,200 competitors) and 2021 (AIcrowd, over 400 competitors). The winning solutions are expected to be usable in plant phenotyping pipelines to robustly assess wheat spike density. We tested this hypothesis by evaluating the 2021 winning solution on an independent dataset consisting of images measured both in the field and in the image by a human, taken with the same acquisition protocol. We use triple collocation analysis to demonstrate that the predicted density appears to be more reliable than the human density measured in the field and in the image. Furthermore, we demonstrate that Global Wheat Head Dataset can be used to estimate wheat ear density from a drone.

Quantifying phenotypes of root-root interactions would allow a greater understanding of how plants react to belowground competition through plasticity of architectural traits. Past research has shown that plants will over proliferate roots in the presence of competition, leaving less resources to allocate above ground, negatively impacting shoot growth and yields [1]. Further evidence highlights a response to neighboring plants in the root architecture of Arabidposis thaliana, as individuals concentrate root mass towards their competitors [2]. To visualize and quantify root architecture plasticity involved in these root-root interactions in real soil, we developed a modified mesocosm system. Within the mesocosm box common bean (Phaseolus vulgaris) seeds were germinated 10 inches apart from each other. Mesh screens were placed on either side of each bean, in order to capture root growth towards each other and/or away from each other. Plants were harvested at the 7-week mark, when the root archetype of each individual was developed and prominent. To harvest our boxes three sides of the mesocosm box was removed and the soil was washed away. Images of each box was taken to produce 3D models. Improvements to the mesocosm system will be made to reliably extract root traits in real soil. These experiments will shine light on an understudied section of crop science and will allow farmers and researchers a better understanding of an otherwise unseen phenomena.

Arsenic (As) pollutes large regions of Asia. Despite phytoremediation initiatives using hyperaccumulators to remove As from contaminated soil, farmers remain reluctant to employ such strategies because of the low biomass and economic value of hyperaccumulating plants. In this study, we demonstrate that cassava can be used for As remediation using a high-throughput root phenotyping platform for cassava roots that we previously developed [1]. Using this phenotyping platform, we identified contrasting root traits associated with As uptake for the two genotypes Rayong 11(R11) and Rayong 90 (R90). Both cassava varieties were grown in pot systems under control (0 mg kg-1 As) and high As (50 mg kg-1 As) conditions and harvested 120 days after planting. We found As stress to reduce shoot and plant dry weight by 57% and 53%, respectively, whereas root dry weight and root traits showed only a slight change. Under As stress, R11 had a 75% higher nodal root number and a 59% lower basal root number than R90. Moreover, R11 root (100 mg kg-1 As) and branch (9 mg kg-1 As) tissues had considerably higher As concentrations than the same tissues in R90. The bioaccumulation coefficient for R11 (2.1) was significantly greater than for R90 (0.9). Additionally, bioethanol yields were unaffected by the presence of As in cassava starch. We suggest that cassava is a promising crop for phytoremediation and that root phenotyping is essential to breed cassava varieties with enhanced As uptake.

Deep learning (DL) methods have transformed the way we extract plant traits – both under laboratory as well as field conditions. Evidence suggests that “well-trained” DL models can significantly simplify and accelerate trait extraction as well as expand the suite of extractable traits. Training a DL model typically requires the availability of copious amounts of annotated data; however, creating large-scale annotated dataset requires non-trivial efforts, time, and resources. This has become a major bottleneck in deploying DL tools in practice. Self-supervised learning (SSL) methods give exciting solution to this problem, as these methods use unlabeled data to produce pretrained models for subsequent fine-tuning on labeled data, and have demonstrated superior transfer learning performance on down-stream classification tasks. We investigated the application of SSL methods for plant stress classification using few labels. Plant stress classification is a fundamentally challenging problem in that (1) disease classification may depend on abnormalities in a small number of pixels, (2) high data imbalance across different classes, and (3) there are fewer annotated and available plant stress images than in other domains. We compared four different types of SSL methods on two different plant stress datasets. We report that pre-training on unlabeled plant stress images significantly outperforms transfer learning methods using random initialization for plant stress classification. In summary, SSL based model initialization and data curation improves annotation efficiency for plant stress classification tasks.

Understanding root traits is essential to improve water uptake, increase nitrogen capture and raise carbon sequestration from the atmosphere. However, high-throughput phenotyping to quantify root traits for deeper field-grown roots remain a challenge. Recently developed open-source methods use image-based 3D reconstruction algorithms to build 3D models of plant roots from multiple 2D images and can extract root traits and phenotypes. Most of these methods rely on automated image orientation (Structure from Motion)[1] and dense image matching (Multiple View Stereo) algorithms to produce a 3D point cloud or mesh model from 2D images. Until now it is not known how the performance of these methods compares to each other when applied to field-grown roots. We investigate commonly used open-source pipelines on a test panel of twelve contrasting maize genotypes grown in real field conditions in this comparison study [2-6]. We compare 3D point clouds in terms of number of points, computation time, and model surface density. This comparison study will provide insight into the performance of different open-source pipelines for maize root phenotyping, and illuminates trade-offs between 3D model quality and performance cost for future high-throughput 3D root phenotyping.

Crop pest detection and mitigation remains an extremely challenging task for the farmers. Majority of the pest classification and detection techniques rely on supervised deep learning frameworks that require significant human intervention in labeling the input data, thereby making the down-stream tasks tedious. Therefore, this study presents a self-supervised learning (SSL) approach to classifying 12 types of agricultural insect pests from 9549 RGB images, by leveraging the Bootstrap your own latent (BYOL) algorithm. SSL uses minimal labeling and is indifferent to data augmentations or distortions. Hence, latent representations from pretrained SSL networks could be generalized well for downstream tasks like classification or object detection. For desirable classification of the insect images, the greatest challenges observed were: i) large intra-class variation (the same insect was found with different colors and patterns), and ii) complex background with inconspicuous foreground. Hence, to overcome these issues and aid generalizability of the representations learned through BYOL, entropy-guided segmentation (segments based on texture not color), is proposed as input to the SSL network in this study. Both raw and segmented images were separately fed to two independent BYOL SSL networks, i.e., with ResNet18 and ResNet50 architectures as the backbone. The efficacy of the latent representations for downstream applications was assessed using linear evaluation, and subsequently compared with traditional transfer learning outcomes from ResNet18 and ResNet50. The results indicated that, while ResNet50 backbone intuitively performed better in all cases, SSL aided with entropy-based segmentation led to ~94% classification accuracy compared to raw images (with ~90% maximum accuracy).

The development of new phenomics approaches to image and process data from the subcellular to ecosystem-scale has accelerated over the past decade. Many of these tools are produced “in-house” within a single lab or a group of collaborating labs, making it hard to keep up with the state-of-the-art for phenomics hardware and software development. The Plant Cell Atlas Phenomics Committee is creating a collaborative space that connects phenomics developers with each other and with the greater plant science community, with the goal of facilitating wide-reaching collaborations. To do this, we will be hosting a video series called “How We Built It” where developers provide a short tour of their inventions and relevant biological applications. We will follow the series with a more in-depth networking event where plant scientists can connect with the inventors and discuss collaborative opportunities. Our goal is to streamline the invention of new phenotyping tools and broaden the application of existing tools.

Multispectral imaging with unmanned aircraft systems (UAS) is a promising high-throughput phenotyping technology that has been shown to help understand the causal mechanisms associated with crop productivity. This imaging technology can accurately predict complex agronomic traits like grain yield within a given generation, creating the potential to fast-track selections in plant breeding and increase genetic gains. The objective of this study was to determine the effectiveness and efficiency of prediction on grain yield in an abnormal drought year across locations within a breeding program. Eleven spectral reflectance indices (SRI) including NDRE, NWI, NDVI, and percent canopy cover were used to evaluate Washington State University winter wheat breeding lines between 2018 and 2021. Data was collected using a DJI Inspire 2 drone, equipped with a Sentera Quad Multispectral Sensor, and collected at the heading date. Lines were observed from single location, single replication preliminary yield trials to multi-location, replicated advanced yield trials. Lines advanced in the breeding program were evaluated across 13 different location-year trials. The calculated SRIs and canopy cover were used individually and in combination as fixed effects in mixed model prediction for grain yield under drought conditions. Models were independently validated with 2021 data. Across locations, SRIs are shown to improve the prediction performance for grain yield under abnormal drought conditions by as much as 40% in the case of NDRE. This research is vital for plant breeders to understand the utility of UAS imaging in variety improvement when dealing with abnormal growing seasons.

The field is not always easy for plant phenotyping using conventional phenotyping platforms due to the limited accessibility and regulated aviation area. Smartphone-triggered ground images were collected on wheat field that has a limited access to monitor growth conditions of four wheat varieties, Shinyoung (SY), Joseong (JS), Taewoo (TW), and Cheongwoo (CW). For field mapping during the growing season, six sets of the raw RGB images were acquired by a smartphone camera in an oblique view angle and processed to transform into nadir view images. A series of algorithms were developed to process the skewed tile images to straighten into the nadir images, align the deskewed images, and stitch them into a field image by detecting crop rows using Hough Transformation. Open-source software, iStitch, was developed to automate the algorithms in a batch process. Plot-level metrics were extracted to analyze plant growth of the wheat varieties using a gridding method for vegetation and leaf area indexes. The processed images resulted in the successful transformation and consistency of algorithms on image alignment and stitching. Plot-level analysis indicated that SY variety performed superior to the other varieties in plant quality and quantity and significantly different from TW variety in canopy coverage. The proposed approach of the stitching and gridding was applied on the skewed images acquired by a smartphone camera but can be directly used for other applications of plant phenotyping on images acquired by a camera on a mobile platform or a grid of stationary cameras in greenhouse or outdoor fields.

Cosegmentation is a recent and rapidly emerging and rapidly growing extension of segmentation, which aims to detect the common object(s) in a group of images. Current cosegmentation methods are ideal and effective only for certain dataset types with limited features that still produce errors making it difficult to obtain detailed metrics of object parts. We propose to build a unified, trainable framework that incorporates multiple features of a high-throughput dataset’s segmented images from multiple algorithms using cosegmentation. Specifically, we propose a novel Cosegmentation for Plant Phenotyping Network (CoPPNet) that utilizes a Fully Convolutional Neural Network with a K-Means Clustering feedback loop for optimal temporal loss. The results from this study will set the benchmark for a novel advancement in computer vision segmentation accuracy and plant phenomics to better understand a plant’s environmental interactions for maximal resilience and yield.

In Canada, winter wheat must survive air temperatures as low as -30oC. Snow cover acts as a thermal blanket reducing the direct exposure to chilling air temperatures. In a joint Can/UK wheat project we measured vegetative cover (VC) on 88 different winter wheat varieties in the fall up to snow cover and directly after the snow melted in the spring. Comparing the VC pre- and post-winter provided a direct measure of winter hardiness. While the majority of the wheat cultivars from the UK experienced greater winter kill than the Canadian lines, there were still some with comparable hardiness. Fall and spring growth rates were determined and the UK lines had similar or greater fall growth rates than Canadian lines but lower spring growth rates. Phenomic determined hardiness was correlated to visual ratings and harvest yield.

Agricultural water resources are threatened by climatic variability and increased competition for available freshwater resources. In order to mitigate the effect of climate change on cotton production, breeders are increasing their efforts on improving drought tolerance in this essential fiber crop. To achieve this, effective screening of diverse germplasm is needed to identify useful genetic variation that can be utilized for crop improvement. Within the last decade, unmanned aerial vehicles (UAVs) have led to the ability to quickly and reliably image large areas while simultaneously decreasing temporal effects associated with a large time window for data collection. This technology allows researchers to scale their phenotyping efforts, enabling studies that utilize mapping and monitoring efforts such as plant water stress detection. In this study, we used UAV-based thermal imagery to screen a diverse population of over 350 different genotypes of cotton in order to locate varieties that exhibit cooler canopies. This diversity panel was grown under two contrasting levels of irrigation, well-watered and water-limited, with data collection flights occurring weekly for three months during the season. The thermal images were clipped to plot boundaries, soil and plant pixels were segmented, and average temperatures were extracted to identify potential drought tolerant varieties. The objectives of this study were to (i) demonstrate that UAV-based thermal imagery, along with our calibration methods, can be used to render accurate plant canopy temperature values and (ii) identify cotton genotypes that outperform others in a drought-stressed environment.

Stomata, microscopic pores on leaf surfaces, regulate the uptake of carbon dioxide and the simultaneous loss of water vapor by leaves. New image acquisition and analysis methods are allowing high-throughput phenotyping of stomatal patterning, which in turn have been applied to better understand the genetic basis of variation in certain species. However, it takes considerable data and effort to train the models, and their ability to accurately detect epidermal structures is constrained to morphologies found within the training data. This issue of context dependency, the inability to perform effectively in novel contexts, is the main hurdle preventing widespread adoption of machine learning in high-throughput phenotyping of intraspecific, interspecific, and environmental variation. Here we show the limited ability of a Mask-RCNN tool, which was previously trained and successfully applied to Zea mays, to analyze images from a closely related grass, Setaria viridis. We then demonstrate successful retraining of the tool to cope with the novel diversity presented by this new species. The stomatal complexes in optical tomography images of mature Setaria leaves were accurately identified by comparison to expert raters (R2 = 0.84). This study highlights the challenge of context dependency for widespread application of machine learning tools for phenotyping plant traits, even in closely related species. At the same time, it also provides a new tool that can be applied to leverage Setaria as a model C4 species, while also providing a roadmap for translation of a machine learning to analyze stomatal patterning in new plant species.

Hyperspectral based prediction of nutrient content in maize leaves Hyperspectral imaging is a promising method to predict crop traits in a high-throughput manner and unlock quantitative genetic studies. A single hyperspectral image can be used to predict several unrelated traits at once using spectral data from 350nm - 2500nm. Researchers have successfully modeled different physiological traits in maize such as vegetative Nitrogen content but the effect of different development stages, genotypes, and treatments on modelling power remains unclear. Here, I explore the ability to model leaf macro- and micro- nutrient content and leaf water content from hyperspectral transmittance data collected with a LeafSpec imaging device. I will compare three different machine learning algorithms; Partial Least Squares Regression, Random Forest and a Convolutional Neural Net to model nutrient content collected from twenty hybrids throughout the 2020 field season in fertilized and not fertilized blocks. Genotypes and development stages excluded from model training are used to externally validate models. Sulfur, Nitrogen, Calcium, Copper, and Iron leaf concentrations were the most amenable nutrients to prediction with coefficient of determination scores from 0.78 - 0.73, respectively. Models trained on samples from a collection of time points were able to accurately predict new time points and genotypes. The findings demonstrate the ability to predict nutrient content in field grown maize over a variety of developmental stages, genotypes, and treatments from a handheld hyperspectral imaging device.

Nitrous oxide (N20) is a greenhouse gas that is three hundred times more potent than carbon dioxide. The majority of N20 emissions worldwide are the result of excess soil nitrogen being metabolized by microbes. It has been hypothesized that crops with better nitrogen uptake efficiency and more roots will reduce excess soil nitrogen therefore reducing N20 emissions. To test this hypothesis, a pilot study was performed in 2021 in collaboration with Iowa State University in which root growth dynamics were captured using RootTracker ™ technology in four commercial maize hybrids. This preliminary study showed a correlation between increased root growth and reduced N20 emissions. Further, we find genetic differences in root growth that is consistent across reps, suggesting that i) cultivar choice impacts N2O emissions and ii) that it is possible to breed for root system architecture to limit N2O emissions. It was also observed that the hybrid with the fastest rate of root growth (lowest N20 emissions) did not reach the greatest soil depth, suggesting early root establishment could be pivotal to more efficient nitrogen uptake. These preliminary results suggest there are differences in root growth by variety that could be exploited to reduce agricultural N20 emissions at scale.

The symbiosis between crops and arbuscular mycorrhizal fungi (AMF) have become an attractive route towards achieving carbon neutral agriculture and reducing the use of chemical fertilizers. Yet, our understanding of how active AMF infections influence the uptake, allocation, and exchange of carbon is limited. Here, we combine X-ray CT and PET imaging to observe and quantify the flow of carbon from leaves to roots to hyphae. Comparison of maize grown with and without AMF allows us to measure changes in the amount of 11CO2 taken up in leaves and subsequently the amount of 11C allocated to below-ground roots. Then, co-registered CT and PET images are used to identify hot spots which may indicate active AMF infection sites. Finally, analysis of 11C kinetics at these hot spots are used to assess the amount of carbon exchanged between maize roots and hyphae. By combining structural and biochemical information, we begin to deepen our understanding of the different types of changes in carbon flow in Maize-AMF systems and how we can improve sustainable agriculture efforts.