A Study On The Extraction Of Leaf Vein In Plant Leaf Image

Plant leaf is an important organ of plant,leaf vein and leaf contour context are two main morphological characteristics of plant leaf,it is of very important significance for the classification and recognition and 3D modeling for plant.In consideration of the problems of current algorithms for extracting leaf vein,including narrow application,low efficiency and result is not satisfactory,a method based on HSV color space and FFCM(fast fuzzy C means clustering method)clustering algorithm using histogram information was proposed.Firstly,the image of leaf is converted from RGB to HSV color space,and then image enhancement and gray-scale morphology processing is conducted.Finally,the FFCM algorithm is used to realize the automatic classification of the blades and the extraction of the leaf vein information.The main contents of this paper are as follows:(1)Sample collection and image preprocessing.As four kinds of plant leaf image recognition and vein extraction are mainly resolved,four kinds of representative samples of leaf images need to be collected.Since the segmentation of the veins requires highly on details,the selected samples should have a certain resolution in addition to their representativeness,and are basically able to ensure the part of the veins clear and relatively complete.The preprocessing of sample images is an important step in this study.In this study,the gray conversion method in HSV color space is used to convert color images into gray images,combined with image enhancement and morphological processing,which significantly enhanced the color difference of mesophyll,leaf vein,and the background,makes the foundation for subsequent extraction steps.(2)Identification of the type of plant leaf images.The current representative leaf vein extraction method based on HSI color space and K-Means clustering solved the problem of extraction on images the intensity of light falling on are even and not even.However,this method is not applicable for leaf images in which pixel values of leaf vein are smaller than that of mesophyll pixel,representative example is yellow leaf images.This study improved the determinant conditions of this method according to the pixel characteristics of this type of leaf images,and redefined the method of computation of the T offset when used on yellow leaf images.Through these measures the new method can be applicable for yellow leaf images,and successfully differentiate and extract the image of leaf veins,which greatly expand the application range of the algorithm.(3)Leaf vein extraction.For images which the intensity of light falling on are even,the leaf vein can be extracted directly by threshold segmentation or clustering.As for images which the intensity of light falling on are not even,the treatment of removing partial mesophyll is needed in order to eliminate the influence of some disturbing pixels on the experimental results.Meanwhile,according to the gray image of yellow leaves,a leaf vein segmentation method on yellow leaves which is based on HSV color space and FFCM was proposed.(4)Improvement of the algorithm.For leaf images with high resolution,method based on K-Means need to deal with large amount of data,and the efficiency is low.To overcome the problem,the FFCM algorithm integrated with histogram information was introduced,which replaced the K-Means clustering used in the previous study of vein extraction.On the basis of not affecting the original extraction effect,the speed of the algorithm was further improved.