凹坑缺陷检测

threeQing

发布于 2019-06-24 15:37:01

1.7K1

发布于 2019-06-24 15:37:01

文章被收录于专栏：机器视觉那些事儿

原图来自Ihalcon论坛

一个非常小的凹坑位于图中间

- 算法思路 -

利用灰度统计特性进行缺陷检测

原图噪声比较大，进行高斯滤波

计算灰度统计特性

intensity (ImageGauss, ImageGauss, Mean, Deviation)

利用灰度统计特性，检测亮缺陷

亮缺陷检测代码如下


*亮缺陷
threshold(ImageGauss, LightRegion, Mean + 3.5*Deviation, 255)
connection (LightRegion, ConnectedRegions)
shape_trans (ConnectedRegions, RegionTrans, 'convex')
inner_circle(RegionTrans, _, _, Radius2)
tuple_find(sgn(Radius2 - max(Radius2) + 0.001), 1, Indices)
select_obj (ConnectedRegions, ObjectSelected, Indices +1)

检测暗缺陷

检测暗缺陷代码

*暗缺陷
threshold(ImageGauss, DarkRegion, 0, Mean - 1.5*Deviation)
fill_up (DarkRegion, RegionFillUp)
connection (RegionFillUp, ConnectedRegions1)
select_shape (ConnectedRegions1, SelectedRegions, 'inner_radius', 'and', 2.0122, 4.9109)
*亮缺陷与暗缺陷相邻很近，根于此关系，选择暗缺陷
select_shape_proto (SelectedRegions, ObjectSelected, SelectedRegions1, 'distance_dilate', 0, 5)

合并缺陷