算法功能：边缘缺陷检测

最新推荐文章于 2024-06-12 10:00:23 发布

胖子工作室

最新推荐文章于 2024-06-12 10:00:23 发布

阅读量597

点赞数 1

分类专栏： Halcon 文章标签：算法

本文链接：https://blog.csdn.net/zhouzongzong/article/details/126651382

版权

Halcon 专栏收录该内容

27 篇文章 1 订阅

订阅专栏

请添加图片描述

*2. 算法思路：
*      （1）初始化
*      （2）使用canny滤波器边缘检测
*      （3）缺陷检测
*           --计算待检测边缘上的点到平滑后边缘的距离，超过一定阈值公差即为缺陷
*      （4）结果显示

*-----------------------初始化--------------------------

dev_update_off ()
*声明空对象，用于保存缺陷边缘
gen_empty_obj(FlawEdgeObject)
*读图
read_image (Image1, 'C:/Users/Administrator/Desktop/1.bmp')

*-----------------------边缘检测--------------------------

*阈值分割+形态学处理，将包含边缘部分图像进行抠图
binary_threshold (Image1, Region, 'smooth_histo', 'light', UsedThreshold)
connection(Region, ConnectedRegions)   
select_shape(ConnectedRegions, SelectedRegions, ['area','anisometry'], 'and', [10000,0.8], [99999,2]) 
closing_circle(SelectedRegions, RegionClosing, 1)
dilation_circle (RegionClosing, RegionDilation, 3.5)
erosion_circle (RegionClosing, RegionErosion, 3.5)
difference (RegionDilation, RegionErosion, RegionDifference)
*抠图处理，得到包换边缘部分图像
reduce_domain (Image1, RegionDifference, ImageReduced)
*使用canny滤波器进行边缘检测
edges_sub_pix (ImageReduced, Edges1, 'canny', 3, 20, 40)
*边缘处理
select_contours_xld (Edges1, SelectedContours, 'contour_length', 10, 2000, -0.5, 0.5)
union_adjacent_contours_xld (SelectedContours, UnionContours, 10, 1, 'attr_keep')

*-----------------------缺陷检测--------------------------
*计算边缘个数，分别进行缺陷检测
count_obj(UnionContours, Number)    

for Index := 1 to Number by 1
    
      *-----算法核心：计算待检测边缘上的点到平滑后边缘的距离，超过一定阈值公差即为缺陷--------
      *选择待检测边缘
      select_obj(UnionContours, ObjectSelected, Index)
      *平滑边缘
      smooth_contours_xld (ObjectSelected, SmoothedContours, 89)
      *阈值公差
      disTH := 5
      *用于保存缺陷边缘X坐标
      flawPtsX :=[] 
      *用于保存缺陷边缘Y坐标
      flawPtsY :=[]
      *得到待检测边缘点坐标
      get_contour_xld(ObjectSelected, Row2, Col2)
      *待检测边缘上点到平滑边缘的距离
      tuple_length(Row2, Length)
      for Index1 := 0 to Length-1 by 1
          distance_pc(SmoothedContours, Row2[Index1], Col2[Index1], DistanceMin, DistanceMax)
          *根据公差判断是否合格
          if(DistanceMin > disTH)
              flawPtsX :=[flawPtsX,Col2[Index1]]
              flawPtsY :=[flawPtsY,Row2[Index1]]
          endif
      endfor      
      *将缺陷点生成区域，并进一步筛选
      gen_region_points(Region1, flawPtsY, flawPtsX)
      dilation_circle(Region1, RegionDilation, 5)
      connection(RegionDilation, ConnectedRegions1)
      select_shape(ConnectedRegions1, SelectedRegions1, 'area', 'and', 200, 99999)
      gen_empty_obj(EmptyObject)
      count_obj(SelectedRegions1, Number1)
      *再从筛选后的区域，选择不合格的边缘
      for Index1 := 1 to Number1 by 1
          select_obj(SelectedRegions1, ObjectSelected2, Index1)
          smallest_rectangle1(ObjectSelected2, Row11, Column11, Row21, Column2)
          clip_contours_xld(ObjectSelected, ClippedContours, Row11, Column11, Row21, Column2)
          concat_obj(EmptyObject, ClippedContours, EmptyObject)        
      endfor
      *连接相邻不合格边缘缺陷
      union_adjacent_contours_xld(EmptyObject, EmptyObject, 50, 1, 'attr_keep')
      
      *合并缺陷
      concat_obj(FlawEdgeObject, EmptyObject, FlawEdgeObject)
endfor
      
dev_display (Image1)
dev_display (FlawEdgeObject)