基于自适应混合颜色空间的图像分割在药片泡罩包装检测中的应用

Application of Color Image Segmentation Based on Hybrid Color Space in Inspection of Color Tablets in Pharmaceutical Blisters

提出了一种建立新的颜色空间方法,根据相关性小和辩识能力好的原则,在经典颜色空间分量中选择颜色分量,构成混合颜色空间,可改善需检测的药片图像分割的精确性和鲁棒性。实验表明:基于混合颜色空间的图像分割,药片边缘定位准确,克服了光照不均匀等问题,可精确计算出药片的几何形态属性,提高缺陷药片的识别率。

A novel method to construct a new color space by selecting a set of color components was introduced which can belong to any of the different classical color spaces. It can improve the accuracy and robustness of segmentation for images of pharmaceutical blister packaging product. Such space, which leads to better discriminating power than classical color spaces, was named hybrid color space. Experimental results showed that edges of tablets are accurate after segmentation based on hybrid color space, and it helps overcome the influence of variable illumination and highlight region. Therefore features of tables like morphological features are calculated accurately, which promotes visual inspection rate.