High-throughput phenotyping of cool-season crops using non-invasive sensing techniques

Chongyuan Zhang

Cool-season crops, including pea, chickpea, canola, and camelina, adapted to a Mediterranean climate, are important crops within several cropping systems. Plant breeding programs have been established with efforts devoted to improving these crops and to developing new cultivars with better performance. However, plant breeding programs still rely on traditional phenotyping techniques that are labor-intensive, time-consuming, and sometimes subjective or destructive. In this dissertation, high-throughput phenotyping or phenomics tools using multiple sensing systems were developed and/or validated to phenotype performance or agronomic traits, such as disease severity, flowering intensity, plant height, and yield. Non-invasive proximal (with handheld device, phenotyping pole, and tractor) and remote (unmanned aircraft systems) sensing platforms with several sensors (e.g. visible/RGB, multispectral, and thermal cameras, hyperspectral sensors, and light and detection ranging sensor) were utilized to collect data. Additionally, custom data processing algorithms were developed to extract features to evaluate or estimate plant traits. In a disease severity monitoring study, there were significant correlations between image-based features (e.g. green normalized difference vegetation index, mean canopy temperature) from multispectral and thermal data with visual rating and yield (|r| up to 0.84 and 0.92, respectively). In a flowering intensity detection study, flowering intensity of bigger flowers, such as pea and canola, can be detected (r < 0.89) with up to 1.1 cm ground sampling distance (GSD), while those of smaller flowers, such as camelina and chickpea, require a GSD of 0.2 cm or better. In a plant height estimation study, moderate to strong correlations between manually measured and proximal (r up to 0.74 and 0.91) or remote (r up to 0.57 and 0.98) sensing estimated plant height were observed. In performance evaluation, the results demonstrated that image-based features (e.g. canopy area) from remote sensing data were significantly (P < 0.05) correlated with yield at early, flowering, and pod/seed development stages, and with days to 50% flowering and physiological maturity at late stages for chickpea and pea. All studies involved demonstrated promising results from sensing technologies to monitor or estimate performance or agronomic traits. With further improvement, these sensor-based tools can assist crop breeding and crop production for rapid crop assessment.

High-throughput phenotyping of cool-season crops using non-invasive sensing techniques

Files and links (1)

Abstract

Metrics

Details