改进的模糊C均值聚类算法和霍夫变换在榛子仁缺陷检测中的应用

Hazelnut Kernel Defect Detection Combined with Improved FCM Algorithm and Hough Transform

以榛子仁为检测样本,采用模糊C均值聚类(FCM)算法进行图像分割;利用飞蛾扑火(MFO)算法改进其目标函数;利用函数对个体样本边缘提取,标记边缘拐点位置,计算拐点个数;对边缘图像进行霍夫(Hough)变换的椭圆曲线拟合,标记并输出饱满籽粒个数;依据试验数据,分析应用改进的模糊C均值聚类算法和霍夫变换对榛子仁缺陷检测的效果。结果表明:改进的模糊C均值聚类算法和霍夫变换,可以准确有效地对饱满、干瘪、霉斑、虫蛀、腐烂的5种榛子仁中的缺陷籽粒进行识别检测,提高榛子仁加工过程中的分拣效率。

With hazelnut kernels as the test sample,the fuzzy C-means clustering(FCM)algorithm was used for image segmentation.We used the moth to extinguish the fire(MFO)algorithm to improve its objective function,extracted the edge of individual samples through the function,marked the position of the edge inflection point,and calculated the number of inflection points.And then we performed Hough transform elliptic curve fitting on the edge image,and marked and outputed the number of full grains.With the experimental data,we analyzed the effect of the improved fuzzy C-means clustering algorithm and Hough transform on the detection of hazelnut kernel defects.The improved fuzzy C-means clustering algorithm and Hough transform can accurately and effectively identify and detect defective kernels in the five types of hazelnut kernels that are full,shriveled,mildewed,moth-eaten and rotten,and the sorting efficiency in the processing of hazelnut kernels can be improved.