关键词: 机器视觉, 石榴, 品质分级, 表面缺陷, 色差分量

Abstract: The method of grading pomegranate appearance quality by manual inspection has low accuracy and efficiency. A pomegranate quality classification method based on machine vision is proposed. Firstly, the pomegranate sample image is collected by machine vision system, denoised and the mask image is obtained. Secondly, the red, green and blue components of the denoised image are extracted, the red and green components are subtracted from the blue component to obtain the color difference image, and the color difference image is segmented by threshold. Then, the boundary of the connected suspected defect region is obtained by mathematical morphology processing, the texture features are extracted, and the defect region is marked according to the different texture features of the defect and nondefect region. Finally, the ratio of defect area to total area and the number of defects are used as the basis for grading, and the quality grade of pomegranate is divided. The experimental results show that the overall classification accuracy of this method is 92.9%, which can effectively and accurately identify the surface defects of pomegranate and classify the quality, and which provides an idea for the industrialization of automatic classification.

Key words: machine vision, pomegranate, quality grading, surface defect, color difference component