Study On The Classification Of Almond Appearance Quality In Hanging Dried

The almonds of hanging dried apricots are one of the nuts rich in many nutrients and health effects,which meets the health requirements of contemporary people.Nowadays,with the improvement of the yield of hanging dried almonds,in the commercialization process is still stuck in the manual or mechanical sorting stage,there are low sorting efficiency,high labor cost input and easy to cause secondary damage and other drawbacks,now with the development of machine vision,because of its high precision,fast speed,non-destructive advantages are widely used in agricultural product testing.In this paper,taking the hanging dried almonds in southern Xinjiang as the research object,the two-dimensional and three-dimensional machine vision technology was used to achieve further research on the hanging dried almonds.In this way,the machine vision grading system is designed in combination with the three-dimensional reconstruction technology to lay a solid foundation for the large-scale application of hanging dried almond commercialization,and its main research contents and conclusions are as follows:(1)Design of almond detection device based on machine vision.Through consulting a large number of materials and research,understanding the commercialization process of hanging dried almonds and the almond sorting standards,according to the relevant national standards,the detection of defective almonds and the sorting of normal almonds are determined as the research direction,and the types of almond defects are damaged and mildew.In view of the shortcomings of the existing grading method,combined with the advantages of efficient machine vision technology and nondestructive testing,the appearance quality inspection device of dried almonds based on the machine vision system mainly includes: conveying device,testing device and sorting device.(2)Almond defect detection based on texture and color characteristics.In the experiment,the color characteristics and texture characteristics are taken as important reference indicators for defect detection,and the characteristic values are extracted after the uniform correction of the surface brightness of the illuminance-reflection model in the HSV model,which avoids the influence of the final judgment result due to the uneven brightness of the almond surface,and combines the fusion of local binary mode and gray scale symbiosis matrix statistics as texture features to achieve defect detection.For the detection of mold in almonds,the R channel value method is used as the basis for whether mildew occurs,and the test finds that the difference between the pixel value and the gray value of the R channel in normal almonds is large,and the average difference value is about 33.4.(3)Study on parameter extraction and grading of almonds based on machine vision.Artificial feature parameter extraction is not related to almond color,in the grayscale of hanging dried almond RGB image,the comparison chose the weighted average method and adopted the median filtering method smoothing process,which can effectively remove external noise while effectively protecting the edge of the image;when looking for and selecting almond features,the octagonal Sobel algorithm is used to calculate the edge of the almond,and the edge is smoother;through morphological operation to eliminate the noise caused by the image segmentation of the hanging dried almond,the closed binary image is obtained.While retaining the contrast of the original color image,the accuracy of the edge detection of the main body of the almond is improved,and the detection value of the physical parameters of the almond is extracted by the minimum rectangular method.(4)Study on the extraction and grading of almond phenotype information based on point cloud.On the basis of extracting the long axis value and width value of almonds,the three-dimensional reconstruction method adds thickness value,surface area and volume as feature parameters.The accuracy of point cloud registration plays a key role in the three-dimensional reconstruction process,after calculating the rotation and translation of the centroid point of the point cloud surface,the mobile least squares surface fitting is used to increase the density of the three-dimensional point cloud,the fitting accuracy is high,and the surface expression is simple.The Poisson curve method meshes the point cloud surface interpolation process to obtain a three-dimensional model of the almond.The almond parameters extracted from the 3D model are combined with the damaged features collected by the 2D image method of almonds based on machine vision,and the BP neural network is built to achieve grading,and the grading accuracy rate is 93%.