Research On Test And Rate Technology For The Wolfberry Appearance Quality Based On Machine Vision

At present,the artificial classification method is widely used to sort different grades wolfberry,which has the problems of low accuracy and productivity,and take much time due to subjective sense.This method makes the appearance quality of wolfberry difficult to be guaranteed and diversity of market demand can not be met.To solve the above problems,computer vision technology has been used to identify and grade surface qualities of wolfberry,which include the size of the wolfberry,the surface defects,and the color.The automatic and non-destructive testing of wolfberry is realized.The main contents and achievements are as follows in this thesis:(1)To extract the image feature for the wolfberry grading detection,analyze and expound the characteristics of several commonly used color space models respectively.At the same time,the image grayscale and binarization,image segmentation,image edge detection,image filtering and so on foundation algorithms of images are studied.Gain the suitable image pretreatment algorithms for wolfberry appearance quality detection and classification.(2)To realize the detection of the size of the wolfberry,the acquire technology for the geometric feature parameters include the area of the wolfberry,diameter of long axis,diameter of short axis,eccentricity,roundness and so on are studied.The correlation model is between the long axis diameter measurements used artificial and the pixel values of the wolfberry is proposed.(3)Identify the defect area based on Hue value,pixels of defect area is extracted after filtering processing.According to the ratio of defect area to the area of perfect wolfberry to discriminate the defect level,the detection and judgment for the degree of surface defect of wolfberry is achieved.The accuracy comes to 93.4%.(4)Extract wolfberry's color characteristic parameters based on HSI color space model,to grade wolfberry by the value of chroma range.Finally,determine the accuracy of classification based on calculate the H,S,I mean,variance respectively.(5)Develop a set of wolfberry automatic detection evaluation system based on the Matlab/GUI software platform.The software adopts the modular design,including file module,image acquisition module,image preprocessing module,characteristic parameters extraction module and the classification module.Realized the unity of human visual operation,friendly interface,simplicity of operator and easy to maintain.Test results prove: using this system to test the wolfberry can not only speed up the classification process,but also improve the accuracy.