Petiole Segmentation Method Based on Multi-Structure Elements Morphology

Abstract

Leaf and petiole can be divided by using computer image processing technology. It has an important role for the feature extracting of leaf and the accuracy of characteristic analysis. In this paper, we compare the distribution of multi-channel histogram in HSV color model, and then select the appropriate channel and threshold value to process the image. Here we use the method of mathematical morphology with the element of multi-structure and multi-scales to smooth the leaf boundary and eliminate the isolated points in the image. Finally, we realize the effective segmentation between the leaf and petiole. Results show that when the condition of the color difference is not obvious, we can better remove the petiole by the use of saturation and brightness channel, and choose the appropriate structure elements according to the direction and size of the petiole. It can provide technical support for the accurate segmentation and extraction of feature information on the leaf.