猕猴桃分级果实表面缺陷的检测方法

A Method for Surface Defects Detection in Kiwi Fruit Classification

猕猴桃的自动化分级中最为复杂、费时的是表面缺陷检测。猕猴桃果实表面缺陷主要包括碰压伤、划伤和日灼,检测过程包括缺陷分割和缺陷识别两个阶段。猕猴桃机器视觉采集系统采用近红外光源采集图像,并对采集图像中值滤波法去除图像采集过程中受到的各种噪声的干扰;图像分析获取最佳阈值,最后图像分割得到猕猴桃果实表面的黑色斑点区域,包括真正的缺陷区域和梗萼区域。通过试验表明,近红外光源能有效提取猕猴桃果实表面的划伤、腐烂伤和日灼缺陷,而且近红外光源图像有效地避免了传统光源图像的反射亮斑区域,通过实验结果,分析针对分割出的可疑缺陷区域如何正确识别,可利用双金字塔数据形式的盒维数快速计算方法,提出描述该区域粗糙度和纹理方向性的特征参数,依此来区分真正缺陷和梗萼区域。

Kiwi surface defects detecting is the most complex,time-consuming in automation and classification.Kiwi fruit surface defect include rotten,scratch and burning,inspection process includes two stages: the defect segmentation and the defect recognition.This paper put forward that kiwi machine vision collection system use the near-infrared light source to get images,and use the median filtering method to eliminate process all kinds of noise of original images;image segmentation get the black spot area of kiwi fruit surface,including the real defect area and fruit calyx area.The experiments indicate that the near infrared light sources can effectively extract kiwi fruit rotten,scratch and burning,and near-infrared light source images effectively avoid the traditional images reflected the light spot area.In the defect segmentation how to correctly identify suspicious defect area,utilizing double pyramid data form of the box counting dimension fast calculation method,put forward the characteristics parameters of describing the regional roughness and texture of the directional,the parameters of the extraction have a little effect by changing light intensity change and fruit position,we can be based on the parameters to distinguish between defects and fruit calyx area.