The Orange Sorting System By Computer Vision

China is a main country of producing oranges, the spead and the accuracy of the orange classification affects the competition of the oranges in the international market. At present, orange classification mainly depends on manual work in China, this method not only needs a lot of person to finish the job, but also has low accuracy and efficiency. Classifying the fruits by using computer vision is the inevitable choice to improve the speed and accuracy of fruit classification. The method of orange classification by machine vision is intelligent with high accuracy and speed and costs lowly, but the large amount of caculation and the complexity of the algorithm are the main problems of this method.This paper introduced the whole orange sortiong system which includes color models, image filtering, segment and edge extraction, the maximum and minimum diameter detection, the detection of image defection,the hardware and software of the system. YUV color model was used in this system to reduce the data quantity; the composition of several color channels was proposed to segment image to improve the problem that the foreground and background of the image can not be segmentted accurately and quickly when it comes to the very complex background. The method of rapid median filtering which has small amount of caculation was used to filter the image and the method of edge extraction by lines was chosen based on the comparation of different methods to reduce the program running time. A new method of diameter detection was proposed based on image segmentation to improve the problem that the old diameter detection algorithms have large amount of caculation and low accuracy. A new method of defection detection which used threshold segmentation combined with regional growth to segment images was proposed in this paper to resolve the problem that the traditional defection detection algorithms have large amount of caculation and long running time. It's proved that the new method was fast and accurate.The software and hardware of the system have been implemented, and the system was used in the actual industrial production. It's proved that the system could sort oranges rapidly and accurately. The expected design goal was achieved.