用活动边界模型精确检测果实表面缺陷

Accurate Measurement of Defect Edge by the Active Contour Models

在利用计算机视觉技术检测果实表面缺陷中 ,为解决泼水算法分割出的缺陷区域比实际大的问题 ,探讨了活动边界模型 (ACM)的原理和求解方法 ,改进了活动边界模型 ,使 ACM在开始收缩时位于目标轮廓外侧 ,且公式能记忆边界的方向 ,当存在两个相距较近的缺陷时 ,ACM会收敛于正确的边界。提出插值算法对得到的不连续边界进行插值 ,从而得到封闭的缺陷边界 。

In order to solve the problem that the defect area divided by the Flooding Algorithm is larger than the real one, this paper is to discuss the basic principles and settlement methods of the Active Contour Models(ACM). The ACM is improved and can be located outside the target edge. And the formula can memorize the orientation of edge. When two defects are close to each other, the ACM will constringe along the appropriate edge. Interpolation Algorithm is proposed to interpolate on inconstant edge in order to obtain close edge. The results show that using the ACM can make the initial edge gained by the Flooding Algorithm more precise. The organic combination of these two kinds of Algorithms can detect accurately the surface defect edge of apples.