Chinese Wolfberry Classification Method Based On Computer Vision

Lycium barbarum L.is a characteristic agricultural product in Ningxia.In 2017,theannual total output value of Lycium barbarum L.is 13 billion yuan.However,its post-harvest deep processing technology is relatively backward.Especially,traditionalmanual sorting has been used in sorting,which has become one of the main reasons for the non-standard development of Lycium barbarum L.industry.The manual sorting method has the advantages of simple equipment,high labor intensity,subjective factors,non-standard grading standards,inaccurate grading,low quality and efficiency,and can not meet the needs of the market.In this paper,machine vision technology is used to analyze and detect the external quality of Lycium barbarum(mainly size,color,etc.).Based on the results,a classification model is established to achieve online detection and classification of Lycium barbarum.The main research contents and results are as follows:(1)Analyzing the current situation of using machine vision technology to study grading detection of agricultural products at home and abroad,and discussing the feasibility and necessity of computer vision detection of Lycium barbarum.Ai ming at the need of grading detection of Lycium barbarum,a sa mple collection and preservation system of Lycium barbarum image was established.(2)After obtaining the Lycium barbarum image,the image preprocessing method suitable for Lycium barbarum quality detection is selected by using the commonly used noise processing and image segmentation methods such as image gray level,Gauss filtering,neighborhood filtering,median filtering,bimodal method and iteration method.(3)In order to eliminate the abnormal color of wolfberry,such as mildew and blackening,which affect the appearance and consumption of wolfberry,the color space of RGB,HSI and HSV was analyzed by choosing appropriate background images,and the color parameters of wolfberry were extracted.Through the analysis of R,G,B,H,S,I,V component histogram,the R component in RGB was selected as the discriminant basis of color defect of Lycium barbarum.The detection rate of abnormal color of Lycium barbarum was 95%.(4)In order to complete the detection and grading of Lycium barbarum size,the projection area,maximum fruit diameter,perimeter and other geometric characteristics of Lycium barbarum were detected.Through analysis,the projection area could be used to more accurately and effectively determine the size and grading of Lycium barbarum.Through K-means clustering analysis and selecting the first,second and third-level samples,the three-level grading benchmark between the projection area and the pixels of Lycium barbarum was trained and established,and the size of Lycium barbarum was detected.The grading accuracy rate reached 96%.Through the comparison of experimental data,the results show that the system has certain practicability.The above research results provide theoretical basis and technical support for the scientificalization and automation of Lycium barbarum appearance quality determination.It has certain theoretical and practical value for promoting the classification and screening of wolfberry and promoting the development of wolfberry trade.