Automatic Plant Leaves Classification Based On Leaf Shape And Vein Structure

Classification of the plant leaves is one important way to study and protect plants and help determine the type of plant and the genetic relationship,there is an important value in the identification of plant and improved variety identification.However,due to the current classification methods exist few categories species,single classification method,unsatisfactory classification accuracy of the large sample. According to the shape and vein structure of the plant leaves,this paper analyzes the blades and207kinds of leaves in leaves library have been researched,The paper mainly work as follows:Firstly,The significance of research topics and research status are introduced.in order to improve the accuracy of the characteristic parameters of the leaf,artificially removing the leaf stalk,and image preprocessing,graying,binarization processing have been used and morphologica-1filtering processing to remove the impact of the holes on the extracted characteristic parameters.Secondly,Extract the contour of the leaf,and calculate the convex hull and smallest bounding box of the contour,at the same time,extract the leaf image feature. Shape aspect use rectangularity axis_ratio,form parameter and invariant moments.vein aspect use energy and contrast and so on,and sort analysis through the shape parameters extracted,at the same time,analysis of the fluctuation interval between leaf species,can classify out the distinct characteristic parameters leaves.Thirdly,the shape factor parameter analysis are used for the rest leaf species.rough classification are processed through the main characteristics of each shape characteristic parameters analyze the corresponding characteristic fluctuation range,combined with other shape features,and it is important way for leaves which shape similarity is high to analysis Mahalanobis distance similarity.and binding the veins structure characteristics about energy,contrast,entropy and inverse difference moment subdivide the leaves.Finally, extract the shape and vein structure features and support vector machine study for Classification about leaves which can not distinguish easily.the experimental results show that the proposed method in the classification of207kinds of leaves,the average correct recognition rate reached91.2%.