基于机器视觉的苹果在线分级

通过CCD相机动态采集苹果两个面的实时图像,提出了泛洪填充+自适应Ostu阈值分割算法提取苹果的轮廓,采用最小外接圆法对苹果上表面图像进行处理得到苹果果径,采用最小外接矩形法对苹果侧表面图像进行处理提取苹果果形特征;将图像进行RGB到HSV空间转换,提取苹果的着色度、果锈,以及疤痕特征,采用基于改进粒子群算法的SVM决策树的分类方法进行苹果的分级。结果表明,该方法对特级果、一级果、二级果和等外果的识别准确率分别达96%,94%,98%,98%,分级速率达4个/s,可以满足苹果在线分级的要求。

Research on apple online classification based on machine vision

Using CCD camera to dynamically collect real-time images of two sides of the apple, a flood filling + adaptive Ostu threshold segmentation algorithm is proposed to extract the outline of the apple. The minimum outer circle method is used to process the upper surface image of the apple to obtain the fruit diameter of the apple. Rectangular method is used to extract the apple''s fruit shape features by processing the apple''s side surface image; the image is converted from RGB to HSV space to extract the apple''s coloring degree, fruit rust, and scar features, and the classification of the SVM decision tree based on the improved particle swarm algorithm Method for grading apples. The experimental results show that the recognition accuracy rates of extra-grade fruits, first-grade fruits, second-grade fruits and other outer fruits have reached 96%, 94%, 98% and 98%, respectively, and the classification rate has reached 4 s 1, which can satisfy the requirement for online-grading apples.