Target Shape Radius-Triangle Description Method And Its Application In The Leaf Image Classification

Plant diversity is an indispensable foundation for terrestrial ecosystems and is closely linked to human survival and development.The reason why human beings can live very healthy and can breathe fresh air on the earth is because plants play a role in purification.Photosynthesis of plants provides an environment for humans to survive.In addition,plants are the largest producers on earth.Photosynthesis of plants uses light energy to produce a large amount of organic matter,providing humans and animals with the energy they need to maintain normal life activities.Plant identification by analyzing leaf image is one of the significant applications of computer vision and image processing.The key issue is how to extract the features that have strong discrimination power for easily distinguishing leaf patterns of different species.In this paper,we propose a new method of shape descriptors based on leaf contours,which is called the Target shape radius-triangle description method.The main research works of this article are as follows:?1?Target shape radius-triangle description method.For each point on the boundary contour line,the radius is first calculated,which is defined as the distance between the boundary contour and the center of the leaf contour.Then,along the clockwise and counterclockwise directions,the length of the boundary arc is equal to the length of the radius to get two points,called the neighboring point.Finally,the contour point is combined with its two adjacent points to form a triangle,which is called a radius triangle,for geometric feature extraction.?2?Using multi-scale research methods.In some vanes with small saw teeth,the use of a single scale sometimes cannot effectively describe its characteristics,which is easy to lose the subtle information features of the vane.Therefore,this paper describes the blade shape from coarse to fine using the multi-scale description method as a whole.?3?Using linear discriminant analysis^[6].In order to improve the blade recognition ability,the high-dimensional feature vector descriptor is further combined with the linear discriminant analysis method to obtain a low-dimensional feature subspace,so as to extract the most discriminative information of the original high-dimensional descriptor.The L₁ distance is used to measure the difference between the two leaf shapes.The method proposed in this paper is mainly used for leaf image recognition and has been tested on five publicly available leaf image datasets including the Swedish leaf dataset.The experimental results show that this method is superior to the most advanced blade image classification methods.In addition,this method has also been tested in the butterfly dataset,indicating that this method is not only suitable for leaf classification,but also has the potential to classify insects.