Research on citrus grading system based on machine vision

In response to the growing demand for standardized fruit quality in the market, a citrus grading system utilizing machine vision technology is designed. The system integrates methods of diameter measurement, colouring rate evaluation, and roundness analysis to achieve precise and efficient automated grading of citrus. In accordance with the national standard GB/T 12947-2008, the grading standards based on the size, colour, and shape characteristics of citrus are established. The citrus image collected in real time is preprocessed using bilateral filtering, grayscale transformation, and Canny edge detection. In order to determine the size, the minimum circumcircle method is used to measure the pixel diameter of the citrus cross-section, and the actual diameter is calculated by the conversion ratio between the actual diameter and the pixel diameter. The proportion of the orange-red area is calculated based on the range of H component in the HSV colour model, and then the citrus colouring rate is determined by combining the flood filling algorithm. Citrus circularity is evaluated by the ratio of the actual contour area of the citrus to the minimum circumcircle area of the citrus. In addition, a graphical user interface (GUI) is created using the Tkinter library to enhance the user experience. Experimental results indicate that the absolute error of fruit diameter is less than 1.5 millimeters, and the colour and shape characteristics of the fruit are consistent with the actual appearance of citrus. The system can effectively and accurately classify citrus fruits, providing technical support for automatic citrus fruits sorting.