The Research Of Rambutan Quality Using Machine Vision

The evaluation methods of color, defects, edible percentage and sugar contents of Rambutan flesh were studied using visible image, X-ray image and CT image.Actuality of Rambutan and agricultural product quality inspecting using machine vision was described. The methods inspecting defects, size, color, stem based on machine vision were explained, and inter quality inspecting of agricultural product based on X ray and CT image was discussed emphasize.The image acquiring devices and experiments were introduced. The image processing and pattern recognize methods for evaluating Rambutan color grade were studied. The Rambutan image was captured using CCD camera. Image background and stem area were removed by OSTU. The color texture feature of image were extracted, sorting model of color were set up base SVM. The correctness of sorting model were 94%,88%,89%,95% for grade 1,2,3,4.The radiograohy course of line-scan detector was analyzed. Then factor of imaging such as reducing noise, choosing suitable voltage and curreny for imaging was studied to improve image quality.A regression model was developed to estimate blank image intensity at higher voltages and cureents. Adusted R2 value was higher than 0.99 for nonlinear fit regession.The prediction method of edible ratio and sugar content were studied base X ray and CT image. X ray image of Rambutan include stem, fluff, shell, flesh, kernel. CT image of Rambutan showed section image of fruit. The shell, flesh, kernel can be seen clearly in image.The segment and fussy C-mean were applied to X ray image in order to get flesh area, the false of edible ratio prediction by X ray image was less 10%. The relation coefficient of model that predict sugar content based linear regression was 92%.A hardware and software system for apple grading was developed. The hardware system can perform Rambutan transport and image capture function; and the software can be used for on line Rambutan grading quality and defect.