基于图像处理的苹果大小分级研究

为了实现苹果的智能分拣,在GB/T 10651-2008《鲜苹果》的标准下,采用图像处理方法对分拣系统中的苹果大小进行分级。首先,针对同一种类的苹果,对摄取到的苹果表面图像进行图像预处理,得到平滑且无噪声影响的苹果二值图;然后,通过寻找每个苹果5幅表面图像中最小外接矩形长宽比最接近1的苹果图像,依此作为获取苹果最大横截面直径的重要依据;最后,通过寻找苹果图像质心,将质心与目标像素点的最大距离作为半径,用最小外接圆的方法得到苹果的最大横截面直径。实验研究表明:选取的30个红富士苹果,最终得到系统分级正确率为93.3%,系统的平均分级时间在1.5 s以内,满足实际苹果分级的需要。所提出的方法可为按照行业标准对苹果进行大小分级提供参考。

Research of apple size grading based on image processing

In order to realize intelligent sorting of apples, under the Chinese standard GB/T10651-2008 Fresh Apples, image process method is used for apple size grading in sorting system. Firstly, the surface images of the same kind of apples are taken and preprocessed, and a smooth and noise-free binary is obtained. Then the image with the smallest circumscribed rectangle aspect ratio closest to 1 in the five images is searched, which serves as an important basis for obtaining the largest cross section of the apple. Finally, by searching apple image centroid, the maximum distance between the centroid and the target pixel is searched, which serves as a radius. The maximum diameter of the apple is obtained by the method of the smallest circumscribed circle. The experimental research shows that for the 30 Fuji apples selected in this experiment, the accuracy rate of system classification is 93.3%, the average grading time is less than 1.5 s, which meets the needs of actual apple grading. The study can provide reference for automatic sorting of apples by size under the industry standard.