Statistical Methodology

Automated estimation of leaf area development

Experimental work in horticulture involves recording and measuring many characteristics of the plants being grown, a process which is known as phenotyping. High-throughput automated plant phenotyping, using data from digital images, has recently become popular, because it offers the potential to make this part of research faster and more efficient.

Leaf area is one important characteristic in understanding plant performance, but it is timeconsuming and destructive to measure accurately. In an EU FP7 funded project, with several partners, including Biometris, BioSS investigated a method for using a histogram of image intensities to automatically measure plant leaf area of tall pepper (Capsicum annuum L.) plants in the greenhouse. During the project, our collaborators created a database of over 15,000 images. Three-dimensional histograms of the distribution of RGB colour intensities were constructed from each image, and the histogram bin counts for pooled combinations of intensities (see figure) were reduced to a small set of principal components that defined the design matrix in a regression model for predicting total leaf area of plants whose leaves were measured manually following destructive harvesting. Regression calibrations were then performed for six different developmental times, enabling predicted leaf areas to be estimated for all plants grown.

The development of leaf area was investigated by fitting linear relations between predicted leaf area and time, with special attention given to the genotype by time interaction and its genetic basis in the form of quantitative trait loci (QTLs). The experiment used several pepper genotypes produced by crossing established but dissimilar cultivars. Although this trial contained a limited number of parental genotypes, the estimation of leaf area from so many plants enabled confirmation of a previously identified QTL for leaf area growth.

Image analysis provides a powerful and efficient way to study and identify the genetic basis of growth and developmental processes in plants. Its use has the potential to make the development of new crop cultivars, as well as any agronomic experimentation to improve plant characteristics, more effective and productive. Further development in this area will benefit from inter-disciplinary work involving agricultural engineering, computer science, statistics and crop genetics.

3D histogram for leaf estimation 3D histogram of colours of pixels from an image of a sweet pepper plant, used to summarise the image as a basis for estimating leaf area. Volumes of spheres increase in proportion to the numbers of pixels represented. Those in an inclined cylinder from lower left to top right contain only the plants, whereas those at the back and shaded blue contain mostly the contrasting background placed behind the plants when images were taken.

Further details from:
Graham Horgan

Article date 2015

Research

Statistical Genomics and Bioinformatics

Process and Systems Modelling

Statistical Methodology

PhD Opportunities

Meetings & Seminars