Discrimination Method Of Varieties Of Rice Based On Machine Vision Technology

Rice is one of the main food crops in China.In the past,the identification of rice varieties generally use artificial methods,mainly by visual inspection method to determine after sampling.Artificial method is characterized by large subjectivity,slow speed,low accuracy.So those methods not only lead to the identification of rice varieties with randomness and low credibility,but also affect people's awareness of rice.Therefore,it is important to identify rice varieties by means of alternative methods.Machine vision refers to use computers instead of people's eyes to "see",to collect and to analyze the image.In the past,because the computer cost is too high,it is so difficult to apply to that the agricultural field still use the original methods.Now with the development of computer cost-performance ratio,computer applications in the field of agriculture has become possible and machine vision is also widely used in the field of agriculture instead of artificial methods.And the artificial method is high subjective,slow speed,low accuracy,while the machine vision inspection has a strong objectivity,high speed,high precision and other advantages.The use of machine vision instead of artificial methods,is the inevitable trend of the development of the times.The paper aims to identify rice varieties using machine vision.The work done is as follows:(1)Collect representative rice samples according to geographical distribution,select whole grains artificially,and collect rice images machine vision system.(2)contrast image preprocessing algorithm such as image edge extraction,image filtering,image segmentation and morphological processing.Then find the image preprocessing algorithm suitable for this topic.The Sobel operator is the most suitable for gray image edge extraction in this paper.Image filtering use median filter algorithm,image segmentation use Otsu threshold segmentation algorithm and binarization image filtering and edge extraction use morphological processing.(3)The related algorithms were proposed and realized by MATLAB programming for detecting the mean value,aspect ratio and circularity of rice.Extract the aspect ratio using the minboundrect function,and achieve automatic detection and error reporting mechanism successful.The experimental results show that the algorithm is reasonable and the three characteristic parameters can be used to identify the rice varieties.(4)A method is proposed to machine classification of rice varieties by using K-means clustering algorithm and BP neural network algorithm.After the machine test and manual test,the test results are as follows:the correct rate of identification of rice and glutinous rice both are 100% and the recognition rate of indica rice is 76%.