基于双边缘模板匹配的类圆颗粒检测

Circle-like Granules Detection Based on Double Edge Template Matching

在各种不同应用场合,经常需要统计类圆颗粒数目,如工业领域的棒材计数、医学领域的细胞分割等,这些类圆颗粒通常形状并非十分规则,且有部分重叠。本文针对类圆堆叠颗粒的二值图像,提出了一种基于双边缘模板匹配的检测方法。首先利用粒度测量方法,得到类圆颗粒的估计半径,然后通过此半径构造一组边缘模板,同时使用Sobel算子将待检测的二值图像分别在两个方向上检测边缘,最后使用构造的模板对边缘进行匹配,并通过限制条件得到预期结果。实验表明,此方法可以有效检测此类颗粒,并且可以解决轻度堆叠、孔洞等影响。

In a variety of applications,it is often necessary to count the number of circle-like granules,such as bar counting in the industrial field,cell segmentation in the medical field,etc.These circle-like granules usually are very irregular in shape and partially overlapped.In this paper,a double edge template matching based detection method is proposed for the binary image of the circle-like stacking granules.Firstly,the granulometry is used to obtain the estimated radius of the circle-like granules,and then a set of edge templates is constructed by using this radius.At the same time,the binary image is detected by the Sobel operators in two directions.Finally,the constructed template is used to match the edges and get the expected results by limiting the conditions.Experiments show that this method can effectively detect such kind of granules and can solve the problems of stacking and holes.