Research On Pummelo's Automatic Detection Index Based On Machine Vision

Quality detection and grading is an important part in the production of pomelo.In view of the current classification of pomelo mainly depends on artificial grading,there are some problems such as low efficiency and inconsistent standards.In this paper,the quality of grapefruit was detected in several aspects based on machine vision,including the fruit diameter,fruit shape,skin color,skin defect,skin texture,peel thickness and edible rate of grapefruit.The relevant detection indexes of grapefruit were quantified and the automatic identification and detection of pomelo characteristics were realized.First,a grapefruit image acquisition box is built,which mainly includes the design of the collection box shape,the selection of lighting scheme,the selection of industrial cameras and lenses.The image of the positive side of the grapefruit is preprocessed,including the acquisition of small area of interest,the comparison of the restoration algorithm of the noise image,the comparative study of the image segmentation method and the research of the image edge detection method.The clear and complete contour of grapefruit is obtained through the custom method.The test system is calibrated.Then,the diameter and shape of grapefruit were detected.The centroid point coordinates of grapefruit are obtained.The maximum inner tangential circle and the minimum outer tangent circle of the grapefruit fruit are obtained based on the centroid point respectively.In order to detect and analyze the size of the fruit diameter of the grapefruit,the method has a good guiding significance for the actual production.The accuracy of fruit diameter detection is verified by designing experiments.Two kinds of grapefruit shape descriptors were put forward in accordance with the requirements of the grapefruit shape in the export standard.The experiment was designed to verify the two descriptors.The results showed that the description of the fruit shape of the grapefruit was more accurate than that of the two descriptors.In order to detect and analyze the boundary contour of grapefruit accurately and intuitively,the grapefruit is placed in the polar coordinate model,and the outline of the grapefruit boundary contour is obtained.This method is intuitive and accurate,and has a high guiding significance for production.Then,the epidermis characteristics of shaddock were detected.Several commonly used color space models are introduced,and transformation between RGB and HSI models is realized,which lays a foundation for studying the color characteristics of pomelo.The skin color and peel defects of pomelo peel were analyzed and detected.According to the color characteristics of grapefruit,yellow and green pixel points and defective partial pixel sets were extracted respectively,and their proportion was calculated respectively.The related characteristic values of Grapefruit's epidermal texture were calculated.Finally,a new method is put forward to detect and analyze the edible rate of pomelo.According to the relationship between the shape and the edible rate of grapefruit,a nondestructive testing model was established preliminarily,which provided a good reference for future research on nondestructive testing model.Based on the expansion principle of the grapefruit boundary contour,the expansion line of pomelo peel was obtained through the transverse and longitudinal images,which provided a theoretical basis for the detection of the thickness of fruit peel of grapefruit.The total volume and section area of fruit and pulp were obtained by image processing.The formula of the edible rate in the standard is derived and converted to make it only affected by volume than one variable.It is well adapted to the fast sampling model of the edible rate of grapefruit.