How to find a specific pattern

For resolution test of production process, a special pattern is inserted. The pattern is pre-defined but the location can vary. The pattern is composed of a few lines (rectagles with various but pre-defined dimensions) and a few characters drawn with rectangles.

How can a ruby script find the location of the pattern?

Comments

  • Hi,

    the solution strongly depends on how the pattern is defined. Could you how a sample?

    Matthias

  • Below is the pattern. Line widths are 26, 28, ..., 35um.

  • I have a proposal:


    # reads pattern pattern = pya.Layout() pattern.read("pattern.gds") # assumes the shapes are on layer 1/0 l1_pattern = pattern.layer(1, 0) # gets the original pattern + merge to normalize areas r_pattern = pya.Region(pattern.top_cell().begin_shapes_rec(l1_pattern)).merged() # reads mask layout layout = pya.Layout() layout.read("mask.gds") # find biggest rectangle of input pattern # (this is assumed to be a good search seed) area = 0 seed = None for s in r_pattern.each(): if s.is_box() and s.area() > area: area = s.area() seed = s.bbox() if not seed: raise("No seed shape found") # computes the pattern's bounding box as the pattern's definition area # (assumes the pattern is rectangular) pattern_box = r_pattern.bbox() # look for seed shapes in the mask layout # assumes the mask shapes are on layer 1/0 l1_layout = layout.layer(1, 0) # select rectangles with the dimensions of the seed dss = pya.DeepShapeStore() # hierarchical mode r_mask = pya.Region(layout.top_cell().begin_shapes_rec(l1_layout), dss) seeds = r_mask.rectangles().with_bbox_width(seed.width(), False).with_bbox_height(seed.height(), False) # now we got the seeds we can check where there is really the pattern we look for for s in seeds.each(): offset = s.bbox().p1 - seed.p1 print("Checking position with offset " + str(offset) + " (DBU) ...") around_seed = pya.Region(layout.top_cell().begin_shapes_rec_overlapping(l1_layout, pattern_box.moved(offset))) # performs a clip to the pattern bounding box clip = around_seed.moved(-offset) & pya.Region(pattern_box) # does an XOR to confirm the pattern here xor = clip ^ r_pattern if xor.is_empty(): # we found the pattern print(" -> CONFIRMED pattern here.") else: print(" -> REJECTED pattern here.")

    The idea is take the biggest rectangle from the pattern, look for the same on the mask and with these candidates do a detailed analysis using an XOR.

    This was my test pattern:

    And this is where the script found it:

    Matthias

Sign In or Register to comment.