摘要
大量密集堆放的烟支的二值截面图像中存在较多的连通区域,为其识别与计数带来了困难。基于模板匹配的方法对烟支进行识别时存在模板中心点定位偏差导致计数不准的问题,本文针对此问题提出了一种识别后的修正方法。烟支具有类圆形几何结构,因此可利用圆间位置关系理论分析烟支间的位置关系,并定义平稳态位置关系。修正的过程便是使烟支的非平稳态位置关系向平稳态位置关系转化的过程。修正后的烟支图像中多计烟支被去除,可达到对计数结果进行修正的目的。
There are many connection areas in the binary section image of densely piled cigarettes, which makes it difficult to recognize and count the cigarettes. The method based on template, matching to recognize cigarettes features the incorrect location of the center of templat, and may cause inaccurate counting. To solve this problem, this paper proposes a method of amending the counting result which is obtained by the recognition method based on template matching. The standard cigarette's geometry structure is similar to a circle, so the positional relationship of cigarettes is analyzed based on the theory of the positional relationship of circles, and then a balanced positional relationship is defined. The amending process inverts the non-balanced positional relationship into the balanced positional relationship. After the amendment, the unwanted cigarettes are subtracted and then the counting result is amended.
出处
《计算机工程与科学》
CSCD
2007年第5期60-61,65,共3页
Computer Engineering & Science
关键词
图像识别
计数
模板匹配
位置关系
image recognition
counting
template matching
positional relationship