Soybean Varieties Identification Technique Based On Computer Vision Research

Longjiang soybean variety is numerous,between different varieties,the quality difference is bigger.Longjiang soybean was mixed up,the supply and demand of Soybean grain are not meet a good docking.Traditional artificial recognition methods exist big workload,and have a low recognition accuracy,grain recognition method based on single feature not accurate recognize classification of soybean grain,and the soybean varieties identification method based on the technology of hyperspectral image is difficult to promote,because the technology and equipment is expensive,complex operation.In order to change this situation,this paper proposes a multiple feature fusion based on computer vision technology of soybean grain of image recognition method.For soybean grain identification based on computer vision technology,the following work is done:(1)Build the soybean image data sets.Gathering NEAU,Heihe river,North feng,Suinong,Heinong,Nenfeng,Hefeng,Kenfeng,Kennong and insect-resistant transgenic soybean and so on,a total of 44 varieties.Each variety selected 150 soybean grains,shoot for single grain,grain of scattered,grain of compact,select and classify for soybean grain images.(2)The image segmentation of soybean.In order to get the expected effect of segmentation,filter for soybean grain images,and segment the soybean grain images after filtering processing,by comparing a variety of segmentation algorithms,found that the threshold segmentation method based on Lab color model of soybean has the best performance of image segmentation.(3)To extract soybean grain image's shape feature,HSI color features and texture features.Then,using principal component analysis method for extracted feature values,which will transform correlation between variables into a set of linear irrelevant variables.By feature correlation contrast experiment,the related values,rectangular,elongation and globular characteristic have small correlation,and then adopting the method of principal component analysis for the rest of the feature value to get the fusion processing,consisting feature space for soybean grain recognition.(4)This paper chooses LWKNCN algorithm to identify 6 kinds of soybean grain images.Comprehensive features on the basis of soybean grain image recognition classification process,the experiment obtain the optimal classification recognition effect when the K value is 2,identify the classification accuracy is 85.61%,the experimental results show that based on feature fusion technology of soybean grain of image recognition technique has high accuracy and strong applicability.This experiment for soybean grain recognition based on computer vision provides a certain reference value,make up the artificial identification of common existence of large amount of labor,the difference of recognition rate of big shortcoming,classification method based on feature fusion technology improve the recognition accuracy,and overcome the based on hyperspectral image classification technology for identification of soybean grain that the technology and equipment is expensive,as well as its complex operation.