一种应用于药品泡罩包装检测的改进分水岭算法研究

Research of Bubble cap Pharmaceutical Packing Inspection System Based on Machine Vision

中文关键词:图像分割  分水岭算法  形态学重建  泡罩药品图像

英文关键词:image segmentation  watershed algorithm  morphological reconstruction  blister drugs image

中文摘要:

      针对分水岭算法存在的过分割问题以及泡罩包装药品图像的特点，提出一种改进的分水岭分割算法对泡罩包装的药品进行分割。该方法将形态学重建运算与使用控制标记符的方法相结合，首先对原图像进行开闭重建，根据邻域像素连通性提取重建后图像的局部极大值作为标记，从而抑制其他的局部极小值，然后依据标记对原始的梯度图像进行修正，最后对修改后的梯度图像进行分水岭分割。试验表明，该方法在缺陷药品检测中，可以有效的抑制传统分水岭的过分割，分割效果良好。

英文摘要:

      Aimed at resolving the problem of over segmentation existing in watershed algorithm and characteristics of blister packaging medicine image, an efficient watershed algorithm is proposed. The method combined morphological reconstruction with control symbol, firstly performs opening and closing reconstruction operations on original image, extracts local maxima of reconstructed image as a marker on basis of the neighborhood pixel connectivity to suppress the other local minima value, then modifies the gradient image according to the marking, and finally, uses watershed segmentation to the modified gradient image. Experimental results show this method has a good segmentation effect, and over segmentation can be significantly reduced.