A statistical approach for analyzing sequences in fruit tree architecture.
Abstract
The comparison of fruit tree cultivars on the basis of their architecture is addressed. We propose to extract sequentially structured data (sequences for short) from collected tree-structured data. Sequences realize a good compromise between the availability of statistical methods (which is the highest for unstructured data i.e. vectors) and the explicit representation of the structural information of plant architectures (which is the highest for tree-structured data). Different approaches are presented for the comparison of fruit tree cultivars relying either on the building of parametric models from samples of sequences corresponding to each cultivar or on the comparison of individual sequences. The strengths and weaknesses of each family of methods, referred either to as parametric or non-parametric, are discussed on the basis of the comparison of seven apple cultivars.