Machine Vision Research Orange Detection And Classification Technology

Detection and classification of navel orange based on computer vision has the vital significance for improve of Gannan and national navel orange market competitiveness. Therefore, this paper studied the navel orange size detection, color detection, defect detection and classification algorithm.The primary research works of this paper are as following:1、Firstly the navel orange image is collected, analyzed the corresponding image preprocessing method, then choose the suitable preprocessing method. At the same time, the navel orange adhesion segmentation algorithm is studied, and worked out that the watershed segmentation algorithm treatment effect is better.2、The size, color,defect detection of the navel orange is studied and the corresponding characteristic value is extracted.①、In size detection, the navel orange is segmented from the background firstly and two size detection algorithms included the bounding rectangle and elliptic regression are put forward.By means of linear regression analysis, elliptic regression method is found better than the external rectangular method, it can better extract navel orange horizontal diameter.②、In defect detection,through counted the R、G and B component of the color information on the Navel orange normal surface and blemish surface,the result when R/B=2and G/B=1.6can roughly extract defect, but exist false detection. Therefore, put forwarded the false detection area recover method by using the ratio of Cb/Cr.At last, the Navel orange of the defect rates have obtained.③、In color detection,this is the first time presented the Colour classification method based the Neural network, and the texture information calculated by chrominance component as the Neural network’s input. The ratio of G and B was used as the threshold value which can detect the green surface, then calculated the Pigmentation ratio of navel orange.3、Four single classification algorithm and comprehensive classification algorithm were summed up based on Navel orange standard。The software of Navel orange detection and classification designed by Matlab GUI, it can effectively complete the task of this subject.4、Testing platform based on DSP was set up to finish capture the Navel orange image and the size, color detect.