Research On Feature Extraction And Classification Of Corn Based On Machine Vision

Corn is one of the word's most widely planted crops as well as main food in China.Nowadays China's arable land reduces year by year,while food demand increases in the circumstance of rapid growth of economy.Therefore,it is imminent to improve food production and among all of it,one feasible and crucial step is selection and cultivation of breeds.In order to breeding process precisely and rapidly,this paper propose a modified multiple threshold segmentation method called hierarchical threshold segmentation.With this method,problem like unideal segmentation can be solved in traditional way of threshold segmentation which has touching corn kernels and large loss in edge part.However,in the approach of this paper,we propose a new calculation method based on corn morphological features including grain size,number of grains and ear row number in picture.This study has realized accurate extraction of corn grain features(grain size,grain number,ear row number etc),research process is mainly divided into two parts:1?The method based on edge detection for corn grain boundary information,which can accurately extract corn size information,its length,width,and aspect ratio.The highest accuracy can reach 97.8%,98.4% and 95.8% respectively.2?Based on proposed multiple threshold segmentation method,corn grain pixel of image histogram is used to determine the corn ear threshold range,and obtain the minimum and maximum threshold,step by step,the method can determine multiple threshold to do clear segmentation of corn ear grain,greatly reduce the adhesion of the traditional segmentation process and successfully separate all corn ear grain.Total grain number detection accuracy can reach up to 98.6%.