Determination Of Surface Defects And Size Of Green Orange Based On Computer Vision

This article studied the detection method of green orange surface defect and the size, which based on the computer vision technology. And research on the following issues:Computer vision system which was suitable for this subject was established and consummated. This system included image acquisition card tycoon, CCD camera, light boxes, mainframe computers, monitors, printers, roller conveyor belts, the three asynchronous motor and NI Vision Development Module Ver8.2.1 software and Lab Windows CVI8.1 which was announced by United States National Instruments.The preprocessing methods of green oranges were analyzed, including:image smoothing, image enhancement, edge detection, image segmentation.Variety of methods were studied which were used to detect the size of green orange, and finally, we chose fruit diameter, the results indicated that the classification accuracy was 93.5%.The grade identification technology of surface defects of green orange based on computer vision was studied. The images of green orange were obtained by CCD camera. After image low-level processing, using National Instruments Vision software, which was produced by America, to obtain the number of the pixels of the fruit diameter, entire fruit, and the defects of green orange, then changed the number of pixels of the diameter to the real diameter, and then calculated the surface areas according to the real diameter. Finally, we calculated the defect areas according to the proportional relationship between surface areas and the number of pixels of entire fruit and defects, and then graded them. The result indicated that the accuracies of four quality grades of green orange were 97.44%,91.49%, 91.78% and 95.12% respectively.