Font Size: a A A

The Study Of Target Recognition Technology In Seed Counting Method

Posted on:2019-02-24Degree:MasterType:Thesis
Country:ChinaCandidate:G N LiuFull Text:PDF
GTID:2393330545470692Subject:Electronic and communication engineering
Abstract/Summary:PDF Full Text Request
As a special commodity,crop seeds play an irreplaceable role in agricultural production.The grain weight of crop seeds is a key indicator of seed size and fullness.It is a measure of seed quality.The measurement of 1000-grain weight is based on seed count,and it is a relatively complicated task to accurately calculate the number of seed particles in a short period of time.Nowadays,the number of crops in our country is mainly counted by manual counting and photoelectric counting,but both of these methods have the disadvantages of low accuracy and high cost.Aiming at this problem,this paper bases digital image processing and object recognition technology.The purpose of crop seed identification and statistics,the study on the algorithm of complete seed number statistical process,and based on the Matlab simulation tools,the adhesion seed count is achieved.Firstly,according to the research background,the corn seed,which is the first in crop yield,is used as the research object,and the preprocessing operation is necessary for the grayscale image and image enhancement;Secondly,this paper introduces several classic image segmentation algorithms,according to the characteristics of the corn seed image and the advantage of algorithm to choose the fuzzy c-means clustering algorithm to coarse segmentation of the image,and the algorithm is proposed for neighborhood gray scale and space feature weighting improvement in order to realize the image seed better separation of target and background;Then,on the depth of still exist adhesion of maize seeds by feature extraction and connected region labeling,will get effective characteristics as the input vector to made up of nuclear parameters have been optimized RBF kernel SVM,complete the construction of the SVM model,so as to realize the recognition of image seed adhesion type;Finally,the seed adhesion type of successful classification can be counted according to the predetermined counting rules,and the accurate counting of seeds can be completed.Through Matlab,a large number of seeds are simulated,results show that the research for statistical number of crop seed particles,counting accuracy is higher,and time-consuming is shorter,the grain index calculation of crop seed particles and has a practical guiding significance for agricultural production field and the general use value.
Keywords/Search Tags:Crop seed counting, Image segmentation, Fuzzy c-mean clustering, Support vector machine
PDF Full Text Request
Related items