一种羽毛虫蛀缺陷检测方法的研究

Research on Defect Detection Method for Wormy Feather

羽毛缺陷复杂多样。一般的图像分割方法很难做到既能适用于多蛀斑分割的复杂情况又能很好地满足工业化羽毛缺陷检测中蛀斑分割的精确性要求。根据毛片特性,提出一种基于机器视觉的羽毛虫蛀缺陷的检测方法;首先采用双边滤波算法对羽毛图像做平滑处理,在有效模糊羽毛自身纹理的同时使得蛀斑边缘信息仍能得到很好地保留,以提高边缘检测的准确性;然后采用一种改进的变分水平集方法进行蛀斑分割,算法的初始轮廓线根据毛片的形状自动生成。实验结果表明这种检测方法分割准确;并能适用于多蛀斑的复杂虫蛀缺陷的检测,较好地满足了工业化检测的需要。

The feather defect is complicated and diversified. It is difficult to segment the multiple feather spots effectively and also meet the demand of detection accuracy preferably with a normal segmentation method in the process of industrial feather wormy defect. An effective method of worm defect detection based on machine vision is proposed according to the characteristics of feather. That is firstly to smooth the feather image with the bilateral fil- ter in order to enhance the accuracy of edge detection. The filter can blur the feather texture and the edge can be well retained at the same time. Then, an improved variational level set method is applied to segment the worm spots. The initial contour can be set against the feather shape. The experimental results show that the proposed method can detect feather worm defect accurately, and even can be used to deal with the multi-spotted feather, so it can meet industry needs preferably.