To study how plants respond to their environment researchers use imaging phenotyping technologies. The use of image-based phenotyping has enabled researchers to analyse plants and produce data at a large scale. However, this large influx of data has created a 'big data' problem to emerge causing researchers to search for new innovative ways to tackle the challenges of processing their data in a reasonable timeframe. To address such issues, deep learning and data science techniques are being used to perform a comprehensive analysis. Here we use a Plant Screen™ compact system to image a series of barley plants using two different imaging sensors. This compact system contains an RGB top and side view camera and a hyperspectral visible near infrared (VNIR) camera. To streamline the processing and analysis of RGB and hyperspectral imaging, we are building a pipeline using a lightweight implementation of the U-Net architecture to improve the accuracy of semantic segmentation based on the raw images captured via the compact system. Several models were designed and developed, each of which was tailored to either the type of imaging sensor being used or the angle for which the images been provided were taken (e.g., top-down, side-view). Results showed that each model regardless of sensor or perspective produced an accuracy greater than 90% and could accurately segment cereal crops regardless of their size, shape or colour. These results demonstrate the feasibility of using DL models to semantically segment cereal crops imaged using either RGB or hyperspectral imaging sensors.

X-ray tomography (XRT) is a powerful and versatile tool for generating detailed non-destructive three-dimensional (3D) image data of large and complicated structures. In particular, excavated, cleaned and dried maize root crowns can be rapidly scanned, and the resulting 3D volumes processed in a computational feature extraction pipeline to provide a wide range of root trait measurements. These measurements provide rich data that give insights into how roots occupy 3D space in ways not possible with any 2D imaging/measurement systems. Hundreds of root crowns can be scanned in a moderate-throughput system, and multivariate statistical analyses can provide valuable insight into the role that genes and quantitative trait loci play in selected root traits. Research presented will describe XRT scan parameter optimization and its impact on root trait data generated by the feature extraction pipeline.

BodyText: Hyperspectral imaging (HSI) system can facilitate the study of crop physiological responses to abiotic stress. It has been established in automated controlled-environment across the globe. Nonetheless, each crop in every new environment requires specific experimental design and data analysis pipeline. At Purdue University's Ag Alumni Phenotyping Facility (AAPF), 15 indica and eight tropical japonica rice genotypes were raised up to 13 weeks old under two nitrogen treatments. HSI data were collected two to three times per week and 14 physiological traits relating to growth, photosynthesis capacity and water transportation were measured manually. With principal component analysis (PCA), physiological trait data showed the effects of subpopulation and treatment whereas only treatment effect could be revealed in HSI data. Changes of reflectance around 715 nm (in the red edge region) were associated with the treatment effect in HSI data based on the loadings of PCA. By training support vector machine classifiers, we found that classification accuracy of treatment levels in HSI data was 80% or greater when the rice plants were six to 10 weeks old. Furthermore, leaf-level nitrogen content (N, %) and carbon to nitrogen ratio (C:N) could be predicted from HSI data by building partial least squares regression models (PLSR) with featured wavelengths. The í µí± ! values for N and C:N were 0.83 and 0.73, respectively, and normalized root mean square error of prediction for N and C:N were 13.67% and 14.39%, respectively (in validation datasets). This is the first study that showed the potential use of HSI on rice at AAPF.

Root exudation refers to the processes by which plants release compounds called root exudates into the soil. These exudates are primarily carbon-containing compounds that interact with microbial communities in the rhizosphere. Microbial consumption of exudates reduces the concentration of the exudated compounds in the soil, causing the plant to exude more of those compounds. Currently, there is limited understanding of the interaction between plant-root exudation mechanisms and the surrounding microbial communities. Among the Sorghum Association Panel (SAP), an established and genetically characterized sorghum diversity panel, we observed a spectrum of root colors (tan, yellow, red, purple-brown, black) identical to the range of observed sorghum seed colors. Previous studies examining differentially expressed metabolites between colorful seeds showed that flavonoids and anthocyanins were higher in dark seeds than white seeds. Root color is genotype-dependent and consistent over time. We hypothesized that the observed color diversity of sorghum roots was due to differential metabolite profiles in the root exudates across genotypes. We designed an experiment to collect exudates from 15 genotypes (n=60). After three weeks of growth, sorghum roots were washed and submerged in ultrapure water for 24 hours. The hydroponic solution was filtered and incubated with methanol. The whole root system was also ground after exudation. The root exudate solutions and the ground-up roots underwent either HILIC and RPLC analysis to separate and detect polar and hydrophobic metabolites. Through metabolite profiling of root exudates, we aim to identify sorghum genotypes that more efficiently allocate carbon below ground via their root systems.

BodyText: A significant portion of plant phenotyping research involves development of new instruments and methodology. Thus, a common experiment is to compare new methods to established ones in order to assess the suitability of the new method. Pearson's correlation coefficient, r, is commonly calculated from the correlation between measurements of the two methods on the same subjects, and it is interpreted to assess whether the new method is a suitable replacement for the established one. However, r (and in this context R, and R 2) is not an appropriate statistic for this purpose, and it provides no meaningful information for comparing quality of methods. This is well established, and other alternatives are known. Here we present quantification and statistical tests of bias and variances of two methods that provide a well-founded approach to method comparison. Comparing newly developed methods to measure height and leaf area index (LAI) using lidar, we find that lidar estimates of height are more precise than established methods and lidar estimates of LAI are equivalent or slightly worse. Using r alone it is not possible to make these interpretations. These sorts of statements are possible due to clear, objective approaches to method comparison, which should be the standard for assessing new phenotyping methods.

Globally, water supply is the major limiting factor for crop productivity. Water use efficiency (WUE) is defined as the ratio of photosynthetic carbon gain relative to water vapor loss from the leaf through the stomata to the atmosphere. Improving WUE would slow crop water use and delay the onset of drought stress when water supply does not meet crop demand. Stomatal density is an important factor that influences plant gas exchange efficiency. We have proof-of-concept that reducing stomatal density in sorghum by ubiquitously expressing a synthetic EPF2core gene can increase WUE without any decrease in photosynthetic carbon gain. However, ubiquitous expression of the synthetic EPF gene has unwanted pleiotropic effects on stem development and seed set. In this study, we test whether tissue-specific promoters can be used to isolate the desired leaf phenotypes without causing unwanted side effects. This provides an important step towards engineering stomatal density to improving WUE and protecting C4 crop yields from drought-induced losses today and in a future, warmer climate.

Community colleges and minority-serving institutions are often ill-equipped with research facilities necessary for providing post-secondary students with opportunities for engagement in authentic research that would equip them with practical skills. Partnerships with research organizations can address gaps in professional training of post-secondary students to better equip them with skills aligned with industry needs, such as data handling. Through collaboration between the Donald Danforth Plant Science Center (DDPSC) and Harris-Stowe State University (HSSU), a local Historically Black College and University, a plant data science Course-based Undergraduate Research Experience (CURE) was developed to improve undergraduate access to research experiences and equip racially minoritized students with cutting-edge data science techniques. Biology and mathematics majors at HSSU used DDPSC researcher-generated plant image data to immerse themselves in real-world plant biology research. Students enrolled in this course used PlantCV, a Python-based image data processing software, to analyze the phenotypes of plants subjected to abiotic stresses. They interacted with image data to investigate heat stress responses in Zea mays (maize) and were exposed to complex interdisciplinary concepts that challenged their understanding of how biology and data science intersect. Students learned to analyze image data to extract conclusions based upon student-generated questions. Qualitative data from students' weekly surveys reveal the building of data science knowledge within students engaged in the CURE, providing insights for educators involved in planning and assessing undergraduate learning experiences at minority-serving institutions. This course

Critical factors that determine crop yields are located underground, making them difficult to analyze. Traditionally, these factors have been measured by growing plants in clear media and measuring traits with visible imaging. Modern phenomics technologies use one or several imaging modalities to capture traits that reflect plant physiology or performance. Analytical techniques for plant phenomics are a crucial part of approaches to achieving desirable agronomic and biological traits. Advances in sensor technologies have paved the way for faster and more efficient plant phenotyping, with methods adapted from disciplines like high-resolution 3D X-Ray computed tomography (CT). A crucial step in their analysis is segmentation-the identification and classification of the scan's voxels as "root" or "non-root". Unlike roots in transparent mediums, roots in non-transparent mediums are difficult to segment from their surrounding materials as root and non-root voxels have overlapping CT values. The challenge we address is the development of neural-driven approaches for volumetric semantic segmentation of plant roots in 3D CT scans, and discuss subsequent trait extraction methods that enable the quantification of root systems and their traits in several agriculturally

High throughput phenotyping and quantitative genetics have enabled researchers to identify genetic regions, or markers, associated with changes in phenotype. However, going from GWAS markers to candidate genes is still challenging. When selecting candidate genes for ionomic GWAS markers, we curated a collection of well-known ionomic genes (KIG) experimentally shown to alter plant elemental uptake and their orthologs in 10 crop species: 2066 genes total. Yet when compared to ionomic GWAS markers, over 90% of significant markers were not linked to a KIG - indicating the list is incomplete and many causal genes are unknown. Continuing to use only functional annotations as candidate selection criteria will keep efforts biased toward well-known genes and hinder the characterization of unknown genes. We propose an unbiased computational approach that compares analogous GWAS markers from multiple species and searches for conserved genes linked to trait markers. Like the KIG list, we expect many of these unknown candidate genes to have orthologs in other species. By leveraging the evolutionary relationship of these conserved genes, rather than prior knowledge and gene annotations, this method: 1) finds more candidate genes than we expect from random chance, 2) selects and prioritizes candidates in poorly annotated species, and 3) includes unknown genes in the results. With this approach, we now have an unbiased list of gene candidates across 19 ionomic traits in model species and crop species to verify in future experiments.

Irrigation of crops accounts for a significant portion of fresh water consumption. In order to utilize this resource more efficiently, it is necessary to engineer crops that can more efficiently use water. Water use efficiency, defined as the ratio of plant growth to water used, is a complex property of plants affected by many different factors. Despite this complexity, genetic variability has been able to be identified in a number of different crops. The C4 model species Setaria viridis remains under-studied in this regard and consequently we sought to identify promising genetic loci contributing to variation in water use efficiency. In order to accomplish this goal we leveraged the high-throughput phenotyping platform at the Donald Danforth Plant Science center to grow S. viridis in well-watered and water-limited conditions. This automated system enables strict control of watering regimes as well as measures of plant traits extracted from photographs using computer vision. Combining these two sets of data allows for direct measurement of whole-plant water-use efficiency on a daily basis which was used as a response variable in a genome wide association study. Significant associations were found for water-use efficiency and related traits. These loci were then prioritized further by pooling information across each day of an experiment and across multiple experiments to zero in on the most likely locations of genes responsible for driving water-use efficiency in S. viridis.

Barley is the most produced crop in Ireland; it is essential as a fodder crop and a key ingredient for the malting industry, contributing to Irish national identity. In Ireland, climate change is bringing more extreme rain; leading to an increase in flooding and waterlogging events. Barley is particularly sensitive to waterlogging urging the need to harness genetic diversity and breed barley with increased waterlogging tolerance to maintain current agricultural production. One of the limiting factors in breeding is the drawbacks of traditional phenotyping. This project relies on modern high-throughput phenotyping, which allow non-destructive, continuous and quantitative data collection. Using imaging sensors in controlled conditions, we phenotyped a collection of barley accessions under controlled and waterlogged conditions using RGB, fluorescent and hyperspectral cameras (VNIR and SWIR). We used the core European Barley Heritage collection (ExHIBiT); made up of 230 diverse 2-row spring barley accessions. This collection was assembled, genotyped, agronomically characterized and its application for association mapping has been established by our team. We observed that 14 days of waterlogging lead to a significant reduction in pixel count (Project Shoot Area) and Quantum yield, showing a large impact on several hyperspectral indices. We also observed that 7 days of recovery after stress are fundamental for differentiate stress resilience. Work is ongoing to optimize hyperspectral image analysis, and establish parameters to distinguish plant performance, enabling to discriminate between resilient and sensitive accessions. These data will be used for association mapping to identify genetic regions contributing to waterlogging tolerance in spring barley.

Jinjeong Lee, Su Jung Ra, Sung Mi Kim, Eunhee Kim, Myeong-Jin Kang,

ORCiD: https://orcid.org/0000-0001-7766-3775 The expanding geographic range of Phyllachora maydis, the fungus that induces Tar Spot infection on corn foliage, is increasingly threatening a Michigan industry that contributes over $1 billion to the state's economy annually. Foliar infection of maize by P. maydis is often difficult to detect early. Visible lesions initially appear tiny, ambiguous, and sparse, making them difficult to identify with the naked eye. Both farmers and breeders of corn desperately need better tools that allow early, definitive detection of lesions and provide more time for management decisions. This tool must verify presence of P. maydis and quantify infection severity as quickly as possible to allow growers the most options for treatment. Advances in machine learning now enable quantification of crop infection presence and severity using powerful object detection packages. With the growing availability of open-source tools, such as the Mask Region-Based Convolutional Neural Network (Mask R-CNN) and PlantCV, the field of plant disease phenotyping has more options for methods than ever before. I propose comparing the accuracy of two potential pipelines to quantify tar spot infection severity: one based on heuristic methods, involving techniques such as dynamic image colorspace thresholding, and the other based on the use of annotations, such as object detection and contour analysis. Comparison of these two methods will provide insight into challenges involved with phenotyping in the field as well as phenotyping foliar diseases using automated methods.

Henry Cordoba-Novoa1, Isabella Chiaravalotti1 , Laura Esquivel1, Robert McGee1, Jeff Smith1, Shangpeng Sun1 and Valerio Hoyos-Villegas1

Ziling Chen1, Jialei Wang2, Xuan Li1, Jian Jin1

Crop breeding relies on the numbers game. The higher the number of locations and entries evaluated, the higher the probability of developing superior cultivars. One major challenge facing breeders of perennial cool season forage crops is the number of biomass harvests per season. In regions with mild winters, alfalfa is harvested every four weeks, six to seven times a year. This creates an operational bottleneck limiting the number of entries and testing locations. Substantial investments were made in the development of automated solutions for precision Ag for row crops. The adoption of these platforms to forage crops rests on their accuracy in estimating biomass yield and cost-effectiveness. This work focused on evaluating the sensitivity of popular unmanned aerial vehicles (UAVs) and imaging strategies for optimal real-time biomass estimates in perennial forage crops. Experimental plots consisting of single plants, row plots, and sward plots were used for a hybrid data collection approach including direct measurements and remote sensing. UAV platforms equipped with a 42-megapixel RGB camera (Sony Alpha 7Rii), a five-band multispectral system (MicaSense RedEdge MX), a hyperspectral sensor (Resonon-Pika L), and a LiDAR (LiDARUSA Revolution 120) were tested. Images were used to generate 3D canopy models of vegetation in the field and to compute morphometric and spectral indices descriptive of vegetation coverage, health and vigor. Harvested biomass yield was used to validate the values derived from UAVs. Preliminary results suggest that the simple red-green vegetation index may be sufficient to give a reliable estimate of biomass yield.

Smart glasses are a rapidly emerging mobile data platform, which can be operated in a hands-free manner through voice commands, a heads-up display and a range of sensors and other digital features. As such, smart glasses enable crop scientists, horticulturalists and agronomists to capture, send and receive digital information, while leaving their hands free to carry out accompanying hands-on tasks or plant manipulations. Phenotypic data increasingly drives agricultural and horticultural development and breeding pipeline discovery. Real-world use cases from innovative agriculture and horticulture technology companies, such as Bayer Crop Science, demonstrate how smart glasses are: 1. serving as a digital phenotyping platform that complements established phenotyping platforms; 2. significantly increases efficiency in phenotypic data collection; 3. facilitate remote collaborations on experiments and other agronomic activities. Smart glass technology integrates easily into existing apps extending capabilities and workflows.

3D scans of real world objects are often represented by point clouds, creating XYZ-coordinates of individual scan points. However, unlike point clouds that are generated from CAD data, points generated from a real world scene lack information about their local context, making segmentation of the structural information contained in the data difficult. Using neural networks (e.g. PointNet) has shown promising results. However, this approach is not well suited for scans of large areas of similar objects, like e.g. a wheat field, because of limitations of the input vector size of the neural network. In addition, point clouds are often unordered, further complicating processing. Since point clouds of biological objects often contain recurring features, we propose to subdivide the point cloud into locally neighboring subsets with a fixed number of points. The collection of subsets can then be used to train neural networks. This approach preserves the original resolution of the point cloud while offering simple data augmentation concepts like creating a number of different subset collections from the same ground truth. There are several advantages to this approach, like significantly simplifying the training phase, because a single, large annotated scan can be sufficient for training, utilizing the similarity of the instances of a plant in the field.