Study On The Vision Detecting Method Of Corn Seed Mold And The Degree Of Mold

Corn is one of the important food sources in our lives.It is one of the top three staple foods in China and an important ingredient in artificial feed for animal husbandry and processing in industry.Therefore,the quality of corn kernels is particularly important.However,it is susceptible to mildew during transportation and storage.Moldy corn kernels affect the survival rate of planting,so the rapid and correct identification of moldy corn kernels is of great significance to reality.In addition,distinguishing corn kernels with different degrees of mildew is also of great significance for further store of moldy corn kernels.The purpose of this article is to be able to detect mold and mildew degree of corn seeds with high accuracy.The paper was carried out from four aspects:discussion on the detection characteristics of corn seed mildew,mold detection method based on R channel,method for detecting mildew degree of corn kernels,experimental verification of mildew detection method and mildew degree detection methodThrough observation,there is a clear difference in color between normal corn kernels and moldy corn kernels.Therefore,based on image processing,four characteristics of gray value,R channel pixel value,G channel pixel value,and B channel pixel value are extracted for for comparative analysis.Divide the corn seed sample into a test set and a verification set,and use the test set to extract the pixel values of the four features.The author counted and analyzed relevant data and found that the pixel values of normal corn kernels and moldy corn kernels on the R channel are clearly defined,and the gray value and G channel pixel values follow it.On the B channel,the pixel values of normal corn kernels and moldy corn kernels are roughly the same,and have cross values,so the R channel pixel value is determined as a feature for detecting whether the corn kernels are moldy.After determining the R channel pixel value as a mildew detection feature,the test set corn seeds are classified into normal corn seeds and moldy corn seeds,obtain the pixel value threshold interval of R channel of normal/mildew corn seeds.Based on this,two methods for detecting mildew are proposed:R channel average method and mildew percentage method.Use the verification set to obtain the detection accuracy of the two methods respectively.The experimental results show that the detection accuracy of the R channel average method is 88%,and the mold proportion method is 100%.Therefore,the mold proportion method is determined to be the corn seed mold detection method.Determine the entire process of corn seed mold detection,including the test set detection process and the verification set detection process.And the verification set detection process,and generate the flow chart.As the degree of mildew deepens,the pixel value of the R channel becomes smaller.Based on this,a method to detect the degree of mildew of corn seeds is proposed:the method of proportion of mildew degree.Determine the R channel pixel value threshold interval of severe/medium/mildly mildew corn seeds.By judging the ratio of the number of severe/medium/mild mildew pixels in the area of interest to the total number of pixels in the area of interest:the proportion of severe mildew Dm1,the proportion of moderate mildew Dm2,and the proportion of mild mildew Dm3.Compare the magnitude of these three variables to determine the degree of corn seed mildew.The BP neural network detection method takes Dm,Dm1,Dm2 and Dm3 four variables as the input layer vector,and gives the output classification result for training and learning.Then use the training and learning results to carry out the classification prediction of the degree of moldy corn in the validation set.Finally,the corn seed mold mildew detection method and the mildew degree detection method are experimentally verified.The experiment obtains the corn seed image through the platform,and obtains the experimental result after analysis and processing by the detection system.The corn seed mildew and mildew degree detection system is developed based on VS C++and Open CV,combined with the mildew detection method and the mildew degree detection method.First select some normal corn kernels and moldy corn kernels as the test set.Through statistical calculations,the R channel pixel value threshold interval,the mold proportion threshold interval,and the R channel mildew degree pixel value threshold interval are obtained.The remaining corn kernels are tested for mildew and mildew degree.The moldy proportion method of corn kernels is used to detect the moldy proportion,and the percentage moldy degrees of corn kernels are tested using the moldy percentage method and the BP neural network detection method.The proportion method of mildew was used to detect corn seed mildew,and the proportion method of mildew degree and BP neural network were used to detect corn seed mildew degree.The experimental results show that the accuracy rate of mildew detection is 100%,which can achieve the purpose of identifying moldy corn seeds with high accuracy.Moldy degree proportion method and BP neural network detection method can classify moldy corn seed samples.