Background Unmanned aerial vehicles (UAVs) equipped with lightweight sensors are making a significant impact in field-based crop phenotyping. UAV platforms have been successfully deployed to acquire phenotypic data in a precise and efficient manner that would otherwise be time-consuming and costly to acquire when undertaken through manual assessment. One example is the estimation of plant density (or counts) in field experiments. Challenges posed to digital plant counting models are heterogenous germination and mixed growth stages that are present in field experiments with diverse genotypes. Here we describe, using safflower as an example, a method based on template matching for seedling count estimation at early mixed growth stages using UAV imagery. Results An object-based image analysis algorithm based on template matching was developed for safflower seedling detection at early mixed growth stages in field experiments conducted in 2017 and 2018. Seedling detection was successful when tested using a grouped template type with 10 subgroups representing safflower at 2–4 leaves growth stage in 100 selected plots from the 2017 field experiment. The algorithm was validated for 300 plots each from the 2017 and 2018 field experiments, where estimated seedling counts correlated closely with manual counting; R 2 = 0.87, MAE = 8.18, RSME = 9.38 for 2017 field experiment and R 2 = 0.86, MAE = 9.16, RSME = 10.51 for 2018. Conclusion A method for safflower seedling count at early mixed growth stages using UAV imagery was developed and validated. The model performed well across heterogenous growth stages and has the potential to be used for plant density estimation across various crop species.

Genomic prediction is becoming a popular plant breeding method to predict the genetic merit of lines. While some genomic prediction results have been reported in canola, none have been evaluated for blackleg disease. Here, we report genomic prediction for seedling emergence, survival rate, and internal infection), using 532 Spring and Winter canola lines. These lines were phenotyped in two replicated blackleg disease nurseries grown at Wickliffe and Green Lake, Victoria, Australia. A transcriptome genotyping-by-sequencing approach revealed 98,054 single nucleotide polymorphisms (SNPs) after quality control. We assessed various genomic prediction scenarios based on Genomic Best Linear Unbiased Prediction (GBLUP), BayesR and BayesRC, which can make use of prior quantitative trait loci information, via cross-validation. Clustering based on genomic relationships showed that Winter and Spring lines were genetically distinct, indicating limited gene flow between sets. Genetic correlations within traits between Spring and Winter lines ranged from 0.68 and 0.90 (mean = 0.76). Based on GBLUP in the whole population, moderate to high genomic prediction accuracies were achieved within environments (0.35-0.74) and were reduced across environments (0.28-0.58). Prediction accuracy within the Spring set ranged from 0.30-0.69, and from 0.19-0.71 within the Winter set. The BayesR model resulted in slightly lower accuracy to GBLUP. The proportion of genetic variance explained by known blackleg quantitative trait loci (QTL) was < 30%, indicating that there is a large reservoir of genetic variation in blackleg traits that remains to be discovered, but can be captured with genomic prediction. However, providing prior information of known QTL in the BayesRC method resulted in an increased prediction accuracy for survival and internal infection, particularly with Spring lines. Overall, these promising results indicate that genomic prediction will be a valuable tool to make use of all genetic variation to improve blackleg resistance in canola.

Automated machine learning (AutoML) has been heralded as the next wave in artificial intelligence with its promise to deliver high-performance end-to-end machine learning pipelines with minimal effort from the user. However, despite AutoML showing great promise for computer vision tasks, to the best of our knowledge, no study has used AutoML for image-based plant phenotyping. To address this gap in knowledge, we examined the application of AutoML for image-based plant phenotyping using wheat lodging assessment with unmanned aerial vehicle (UAV) imagery as an example. The performance of an open-source AutoML framework, AutoKeras, in image classification and regression tasks was compared to transfer learning using modern convolutional neural network (CNN) architectures. For image classification, which classified plot images as lodged or non-lodged, transfer learning with Xception and DenseNet-201 achieved the best classification accuracy of 93.2%, whereas AutoKeras had a 92.4% accuracy. For image regression, which predicted lodging scores from plot images, transfer learning with DenseNet-201 had the best performance (R2 = 0.8303, root mean-squared error (RMSE) = 9.55, mean absolute error (MAE) = 7.03, mean absolute percentage error (MAPE) = 12.54%), followed closely by AutoKeras (R2 = 0.8273, RMSE = 10.65, MAE = 8.24, MAPE = 13.87%). In both tasks, AutoKeras models had up to 40-fold faster inference times compared to the pretrained CNNs. AutoML has significant potential to enhance plant phenotyping capabilities applicable in crop breeding and precision agriculture.

BackgroundWithin the Brassicaceae, six species from the genus Brassica are widely cultivated throughout the world as oilseed, condiment, fodder or vegetable crops. The genetic relationships among the six Brassica species are described by U’s triangle model. Extensive shared traits and diverse morphotypes among Brassica species make identification and classification based on phenotypic data alone challenging and unreliable, especially when dealing with large germplasm collections. Consequently, a major issue for genebank collections is ensuring the correct identification of species. Molecular genotyping based on simple sequence repeat (SSR) marker sequencing or the Illumina Infinium Brassica napus 60K single nucleotide polymorphism (SNP) array has been used to identify species and assess genetic diversity of Brassica collections. However, these methods are technically challenging, expensive and time-consuming, making them unsuitable for routine or rapid screening of Brassica accessions for germplasm management. A cheaper, faster and simpler method for Brassica species identification is described here.ResultsA multiplex polymerase chain reaction (MPCR) consisting of new and existing primers specific to the Brassica A, B and C genomes was able to reliably distinguish all six Brassica species in the triangle of U with 16 control samples of known species identity. Further validation against 120 Brassica accessions previously genotyped showed that the MPCR is highly accurate and comparable to more advanced techniques such as SSR marker sequencing or the Illumina Infinium B. napus 60K SNP array. In addition, the MPCR was sensitive enough to detect seed contaminations in pooled seed samples of Brassica accessions.ConclusionA cheap and fast multiplex PCR assay for identification of Brassica species in the triangle of U was developed and validated in this study. The MPCR assay can be readily implemented in any basic molecular laboratory and should prove useful for the management of Brassica germplasm collections in genebanks.Electronic supplementary materialThe online version of this article (doi:10.1186/s13007-017-0200-8) contains supplementary material, which is available to authorized users.

Two stilbenes, resveratrol and pterostilbene, exhibit antifungal activity against Leptosphaeria maculans, the fungal pathogen responsible for blackleg (stem canker) in canola (Brassica napus). In vitro studies on the effect of these stilbenes on L. maculans mycelial growth and conidia germination showed that pterostilbene is a potent fungicide and sporicide, but resveratrol only exerted minor inhibition on L. maculans. Cell viability of hyphae cultures was markedly reduced by pterostilbene and SYTOX green staining showed that cell membrane integrity was compromised. We demonstrate that pterostilbene exerts fungicidal activity across 10 different L. maculans isolates and the compound confers protection to the blackleg-susceptible canola cv. Westar seedlings. The potential of pterostilbene as a control agent against blackleg in canola is discussed.