Research And Development Of Apple's Quality Grading Technology Based On Machine Vision

As a fruit country,China’s annual output of fruit has always been among the highest in the world.A very important part of fruit sales before it is fruit grading.If the harvested fruit is graded,the value of the fruit can be improved.However,China is not as advanced as foreign countries in the field of automatic fruit grading.Most of the methods used in fruit grading in China are manual grading and mechanical grading.There are many deficiencies in these two grading methods: low grading efficiency,easy damage to fruits,and inconsistent grading standards.Therefore,this paper selects the apple with the first fruit yield in China as the grading target,and uses the machine vision technology to study the hierarchical detection algorithm of apple color dimension,size dimension and defect dimension,and simulates and verifies the algorithm.This paper first establishes an image acquisition system and performs preprocessing operations on Apple images,including image graying,image denoising,image segmentation and morphological processing.Then,the extraction of the apple size and color features is performed,and the statistics of the relevant hierarchical data are completed.In terms of apple color,the RGB color space model is converted into HSI model,and the surface color grade is divided by calculating the red coloration rate of apple;in the aspect of apple size feature extraction,the apple diameter is calculated by the minimum circumscribed circle method.Then the external features and support vector machine are used to identify the apple defect.Firstly,the image segmentation algorithm is used to threshold the image.The closed contour,edge extraction and subtraction are used to obtain the outer contour of the apple calyx,fruit stem or defect,and then the hole is filled.Separate from the connected area and finally split the area of interest.Using color features,texture features and geometric features,after extracting 13 eigenvalues,the support vector machine is used for recognition.The accuracy rate is up to 91.32%,which realizes the grading of apple defect dimension.Finally,the convolutional neural network algorithm is introduced into the apple defect,and the 8-layer convolutional neural network is designed.The apple defect is used for identification.The accuracy is up to 96.89%.The difference from the artificial grading is small,meets the grading requirements,and uses the new verification.Compared with the SVM and CNN algorithms,CNN is superior to SVM algorithm,and CNN does not need to extract external features.