Research Of Dried JunJuJube Size Classification Method Based On Machine Vision

This paper established a computer image acquisition system,through the camera to obtain images of the dried jun jujube,and then to grayscale image,image enhancement and image processing,image binarization,image morphology filling and processing,such as pretreatment algorithm in image segmentation from the morphology and surface texture features of dried red jujube extract quickly.Based on dried jun jujube in the region of the disease and the disease disease area is extracted tonal value difference,the ratio of surface area to extract the disease area and jujube as a threshold to determine disease area identification accuracy,combined with the feature of dried jun jujube region color value H mean and mean square error,using SVM method to set up the identification model of dried jun jujube diseases,Otsu for dried jun jujube areas’ s threshold segmentation,image statistics of local properties and morphology processing,extraction of crack in binary image,based on the crack image from constant method to establish the crack identification model,through the analysis of experimental results,found the bird pecked and phytotoxicity of low detection accuracy,Only 86.7% and 76.7%,respectively,the detection accuracy of mildew can reach 100%.Using dried jujube maximum distance between two points on the contour search algorithm is used to measure dried jun jujube longitudinal diameter,using the least squares linear union the diameter of the pixel values of the real value and function relation of dried jun jujube size classification model is established,and compared with the grading standard,so as to realize dried jun jujube size grading,will eventually actual measured results with the model,comparing the measured result was designed based on GUI interface dried jun jujube size grading system,classification accuracy is 95.3%,the results reached the anticipated design requirements.