Research On Potato Grade Classification Based On Machine Vision

Among the four major foods in the world,the potato is called food that can save mankind by experts.Its distribution area is very wide,but the intelligent sorting of potatoes is still in the researchful stage,and many areas still use the manual sorting method.During production,the qualities of potatoes will seriously affect the actual benefits of the potato deep processing industry,so the quality classification of potatoes is one of the important prerequisites for industrialized production.Most of the existing mechanical grading methods are prone to cause secondary damage to the surface of potatoes.In view of the current status,it is particularly important to develop a rapid grading method that does not damage the qualit ies of potatoes.In this paper,a quality grading system of potatoes based on machine vision technology is designed to meet the potato classification requirements in terms of surface defects,shape,and quality.While ensuring the accuracy of potato quality classification,multiple potato samples information ared collect on the same image.In turn,improves the efficiency of classification.To this end,this paper designs a classification system of potatoes based on vision technology,including the following aspects:1)For the complete extraction of potato es surface information,a smooth transparent plate is used as the potato placement platform,and the RGB-D camera is used to collect the potato the complete surface information by twice obtaining.Aiming at the image preprocessing link,actual comparison of different algori thms in each link,analysis of their effects,and determine the best processing plan.In this paper,median filtering method is used to denoise,and canny operator is used for extracting edge information.Histogram threshold method is used for segmentation,and morphological opening and closing operations are used to prepare for the following defect detection and shape classification.2)Since surface of potatoes defects have obvious color characteristics,a method based on HIS color(hue,saturation and bri ghtness)model is proposed to detect various external defects that are likely to exist on the surface of potatoes.For the plaque,the RGB model is used.Using two color models(HIS and RGB),the detections of different types of defects in potatoes are realized,then finally the defect area is marked.3)To research the application of surface image information in shape classification of potatoes,this paper uses the obtained surface features of potatoes as the relevant parameters of support vector machine(SV M)for training to achieve the purpose of shape classification of potatoes while improving the accuracy of recognition rate.In this paper,the shape of potatoes is roughly divided into two types:spherical and deformed,with an average classification accu racy of 95%.4)Different from most existing quality detection methods of potatoes,this article uses RGB-D camera to collect RGB and IR images of one half of the sample potato es,combined with 3D point cloud reconstruction technology,and performs of potatoes point cloud reconstruction.Use curvature constraint method,inverse distance interpolation method,and voxel grid method for the preliminary data processing work.And finally performs dense reconstruction based on the area method of the patch model.Use the tetrahedral grid structure method for volume according to calculation.The density of potatoes is generally stable at 1.0-1.2g/cm~3,then the quality of potato samples can be calculated to realize the quality classification of potatoes.