Application of image analysis for grass tillering determination

Abstract

Tillering is defined as the process of above-ground shoot production by a single plant. The number of grass tillers is one of the most important parameters in ecology and breeding studies. The number of tillers is usually determined by manually counting the separated shoots from a single plant. Unfortunately, this method is too time-consuming. In this study, a new method for counting grass tillers based on image analysis is presented. The usefulness of the method was evaluated for five grass species, Phleum pratense, Lolium perenne, Dactylis glomerata, Festuca pratensis and Bromus unioloides. The grass bunches were prepared for analysis by cutting and tip painting. The images obtained were analysed using an automatic procedure with separation of shoots and other objects based on morphological parameters. It was found that image analysis allows for very quick and accurate counting of grass tillers. However, the set of morphological parameters for object recognition must be selected individually for different grass species. This method can be recommended as a replacement for the traditional, time-consuming method in grass breeding.

Keywords: Grass; Image analysis; Morphology; Tillering.

Fig. 1

Grass bunch before (a) and after (b) cutting. 1—grass stem cut at the 5 cm height, 2—white painting roller, 3—digital camera for capturing the image of the bunch

Fig. 2

Flow diagram of the algorithm: a original image after cutting and painting; b image of the grass bunch after splitting to RGB bands (blue band left); c segmentation, red colour—object of interest, black colour—background; d image after morphological opening with small objects removed; e image after filtering according to the selected morphological parameters