Extraction Of Leaf Vein Based On Canny Algorithm And Hue Information

Leaves of plants reflect the plant species and the growth state of the plant directly.The leaf veins contain important physiological information of the plant,so that the extraction of the veins is the key of plant modeling and identification.With the rapid development of computer image processing technology,it has been widely used in plant research.The establishments of databases of plants and research on vein extraction have attracted the attention of people.But the leaf veins are complicated or hardly distinguished,so that existing vein extraction methods usually could not get the satisfied extraction effect.This thesis focuses on leaf vein extraction,and proposes a new method of leaf vein extraction that combines Canny algorithm and hue information.Furthermore,the morphological pre-processing and wavelet transform technology are applied to the research of leaf vein extraction.The main work of this thesis is shown as follows:(1)Compared with several existing traditional edge detection algorithms,we choose Canny algorithm for leaf vein extraction.First of all,several traditional edge detection algorithms are compared with the analysis of their respective advantages and disadvantages,then the principle of Canny algorithm and its three evaluation criteria are analyzed.Secondly,Canny algorithm is used to extract the leaf veins of 32 different types of leaves from the Flavia dataset.Finally,the experimental results of several different edge detection algorithms are compared.(2)Using the hue information of HSI color space model to extract leaf vein.First of all,compared with several commonly used color space,the hue component of HSI color space is used to extract leaf vein,and the process consists of three steps,i.e.,color space conversion,image enhancement and binarization.In the end,the experimental results of different kinds of leaves are compared.(3)Mainly,apply morphological pre-processing to the leaf vein images obtained by the above two methods and then fuse the results.Firstly,dilation and erosion are used to process the leaf vein images obtained on Canny algorithm and hue information above respectively to eliminate noise and fill the fracture.Secondly,the results obtained on these two methods with the wavelet transform are fused.Different fusion rules are selected for the high frequency coefficients and low frequency coefficients,and the optimal wavelet decomposition level is selected through experiments.Finally,the effect of the fusion image is evaluated by three objective criteria,including information entropy,average gradient and mutual information.(4)Evaluate and analyze the results of fused vein extraction.First of all,the standard leaf vein images are determined.Then,the method of leaf vein extraction proposed in this thesis is evaluated by several objective criteria,including misclassification error,root mean square error,peak signal to noise ratio,correlation coefficient and distortion.The experimental results show that the proposed method can effectively improve the accuracy of vein extraction.