figure[1]Figure 0.table[1]Table 0. The last few years have seen an increasing interest in the development of high throughput phenotyping platforms (HTPP) that allow the automated measurement of plant growth and structure. These platforms have adopted various imaging technologies, such as fluorescense imaging, thermal imaging, visible imaging and so forth. Since plants are morphologically complex and inherently three dimensional, 3D imaging and reconstruction approaches have obvious advantages over 2D analysis or over any other traditional methods in terms of high temporal and spatial resolution. Besides, it provides a non-destructive alternative to traditional methods with a cost effective approach. By using the information generated by 3D phenotyping platform, researchers are able to investigate complex traits related to the growth, responses to external and internal signals or perturbations.

At present, 3D reconstruction for phenotyping platforms mainly focus on using 3D laser or LiDAR scanning for large scale plants in the field. Image-based reconstruction, on the other hand, has the advantages of being cost-effective, in general, it is able to perform well on complex structures and small plants as well as retaining the colour information. Therefore, it’s well-suited for phenotyping in indoor or controlled environments.

In this paper, we present new computational approaches applied to analysing the growth of Arabidopsis thaliana rosettes over time from stereo reconstruction. By applying a semi-global matching algorithm,2.5D point-clouds of (full) Arabidopsis trays were generated. The point-cloud was filtered and then segmented in order to isolate automatically each single plant by clustering methods. Post-processing of the point clouds generated meshes were using various surface reconstruction methods. These meshes were analysed to quantify various attributes of plants, such as plant dimension in three dimensions, surface area or the dynamics of plant development over time. The overall approach proved useful in quantifying precisely morphometric parameters in 3D for a set of Arabidopsis accessions and in deriving supplementary phenotypic informations from the the initially-collected datasets.