Font Size: a A A

Research On Algorithm Of Feature Extraction And Recognition Based On Edge Geometry Of Leaves

Posted on:2019-04-18Degree:MasterType:Thesis
Country:ChinaCandidate:X Y XuFull Text:PDF
GTID:2370330548963489Subject:Engineering
Abstract/Summary:PDF Full Text Request
In all life forms on the earth,the plant life form is the largest and most widely distributed,both in quantity and variety.The existence of plants has a huge impact on the ecological environment in which we live,and it is indispensable.Without plants providing food for us,there will be no agriculture and no modern civil society.Agriculture is the lifeblood of modern society and economy,and it is the basis for every human being to survive.Therefore,research on plants has been ongoing.For a wide variety of plants that are widely distributed,it is necessary to study their classification and identification work.It is also a tedious task.With the advancement of machine and computer technology,the use of digital automation can quickly promote the study of plant classification,which can help us to better understand the plant,make more plants available to humans,and better protect Endangered and rare plants.The use of computer classification system can be better applied to the collection,extraction,classification and identification of plants.In all living things on the earth,the plants are considerable in quantity and distribution,and they have a tremendous regulatory effect on the ecology of the entire earth.They are an important guarantee for human survival and prosperity.At the same time,not only plants produce oxygen for human,but also provide humans with a large amount of carbohydrates,which are huge treasure for human survival and development.As the plants are accompanied by humans,they will also be destroyed by the destruction of human beings.Some plants will disappear from the earth forever.Therefore,in order to protect the earth's plant resources more effectively,humans need to better understand plants and study their different taxonomic habits,which has also an important practical significance for the study of plant classification.Only by fully understanding the plants can we better protect and utilize them.Nowadays,the development of computer classification technology can be well used in the classification and identification of plants.In the composition of the plant are mainly: roots,stems,leaves,flowers and fruits.As a result,they both morphologically and structurally mutate.Therefore,they will have different degrees of difference in appearance,size,height,color,etc.This determines that it is not possible to use images of roots or stems to identify plants.Other parts of the plant,such as flowers,leaves,fruits,and seeds,are relatively stable and will not undergo major changes.In this way,their image data will be more suitable for processing with a computer.However,the flowers,fruits,and seeds of plants are mostly three-dimensional shapes,which makes the recognition more complex and cumbersome.The leaves of plants are important organs of plants for photosynthesis and nutrient production.They have the features of easy collection,stable characters and obvious characteristics,so they can be used as a method for plant identification.Leaf recognition is a type of pattern recognition.This technology is widely used in the remote sensing telemetry,the judgment of crop growth,the forest vegetation,and the species judgment.At present,the main solution adopted for leaves recognition is to use BP neural network,fuzzy algorithms,etc.Most of these solutions are focused on identifying leaves from the perspective of the image.These algorithms extract the leaves feature values by extracting the contours of the leaves edges for image analysis or according to the structure of the leaves edges and the total combination of the conditions.This method not only has a large amount of calculation,but also needs to extract feature values from the whole leaf,and the efficiency is low and the recognition rate is not high.The scheme adopted in this paper is to analyze the processing and classification of leaves from the perspective of time domain electrical signals.Extract the contour signal from the two-dimensional spatial feature image of the edge of the leaves,and map to a 2D plane coordinate system.This coordinate system can be a rectangular coordinate system or a polar coordinate system.Through experimental analysis,the blade edge feature extraction scheme in cartesian coordinate system is characterized by good stability,simple data processing,and intuitive visualization.However,the dimension of feature data is relatively high,and the amount of calculation of some blades is very large.These are important fatal weaknesses.This determines that the cartesian coordinate system is more suitable for leaves with sharp leaf blade ripping and sharp blade jagged.The feature values of the edge extraction performed in the polar coordinate system can be processed into a one-dimensional array.It is only necessary to traverse these one-dimensional array data to obtain the feature data.It can be seen that the amount of computation will be relatively small,and the speed will be very fast.Therefore,the extraction method in the polar coordinate system is more suitable for the leaf edge change is relatively flat,the leaf crack is not obvious and the leaf character is similar to a round or oval shape.Compared with the extraction method in cartesian coordinates,it can reduce a lot of calculation time.
Keywords/Search Tags:Leaf recognition, Leaf edge feature extraction, The maximum peak valley ratio, Surface area ratio
PDF Full Text Request
Related items