項目需求：要求圓環外圈毛刺的凸出高度不得超過20像素，如超過判為NG。
算法分析：按照一般的思路，應該是先將外圈擬合成一個圓；然后將這個工件閾值化，提取工件區域；然后用該區域減去之前擬合的圓，剩下的區域就是毛刺部分了。那么什么樣特征的毛刺算超標的呢？很顯然，如果僅僅用'area'、'inner_radius'等參數去篩選，至少從原理上來看，是不完全契合項目需求的。
具體完整算法如下：
1 read_image (Image, 'C:/Users/02/Desktop/Ring.png') 2 dev_set_draw ('margin') 3 threshold (Image, Region, 0, 253) 4 fill_up_shape (Region, RegionFillUp, 'area', 1, 20000) 5 fill_up (RegionFillUp, RegionFillUp1) 6 difference (RegionFillUp1, RegionFillUp, CircleInner) 7 smallest_circle (CircleInner, Row, Column, Radius) 8 find_circle (Image, PartCircleXLD, Regions, Cross, Circle, Row, Column, Radius + 100, 0, 360, 40, 100, 20, 1, 40, 'negative', 'first', 'inner', 10, 'circle', RowCenter, ColCenter, Radius1) 9 gen_region_contour_xld (Circle, RegionCircle, 'filled') 10
11 *稍微膨脹一下，便於后續處理 12 dilation_circle (RegionCircle, RegionCircle, 1.5) 13 difference (RegionFillUp, RegionCircle,RegionDifference) 14 opening_circle (RegionDifference, RegionDifference, 3.5) 15 connection (RegionDifference, ConnectedRegions) 16 select_shape (ConnectedRegions, SelectedRegions, 'area', 'and', 50, 9999999) 17 count_obj (SelectedRegions, Number) 18
19 *后面精確測量一下多出的毛刺是否超標 20 gen_empty_obj (RegionDefects) 21 if (Number >0) 22 for i := 1 to Number by 1
23 select_obj (SelectedRegions, ObjectSelected, i) 24 distance_pr (ObjectSelected, RowCenter, ColCenter, DistanceMin, DistanceMax) 25 *如果毛刺凸出高度超過20像素，則判斷超標 26 if (DistanceMax - Radius1 > 20) 27 concat_obj (RegionDefects, ObjectSelected, RegionDefects) 28 endif 29 endfor 30 endif 31
32 dev_display (Image) 33 dev_display (RegionDefects) 34 disp_cross (3600, RowCenter, ColCenter, 20, 0.45)
與程序中的抓圓函數find_circle相關的函數，可以從這里下載：
https://files.cnblogs.com/files/xh6300/find_line_find_circle.zip