X-Git-Url: http://git.tomasm.cz/imago.git/blobdiff_plain/e4044c7515289bb999ee98698abbe09aa6e4b517..bc8264a583025d7be3e3238aa40d4950050f7fcf:/ransac_grid.py

diff --git a/ransac_grid.py b/ransac_grid.py
index 3e06002..a483a9e 100644
--- a/ransac_grid.py
+++ b/ransac_grid.py
@@ -3,6 +3,7 @@ import matplotlib.pyplot as pyplot
 from math import sqrt
 import random
 import sys
+import time
 
 import src.linef as linef
 import src.gridf as gridf
@@ -10,6 +11,8 @@ from src.manual import lines as g_grid
 
 import new_geometry as gm
 
+random.seed(12345)
+
 def plot_line(line, c):
     points = linef.line_from_angl_dist(line, (520, 390))
     pyplot.plot(*zip(*points), color=c)
@@ -27,8 +30,29 @@ def to_general(line):
 def nearest(lines, point):
     return min(map(lambda l: dst(point, l), lines))
 
+def nearest2(lines, point):
+    return min(map(lambda l: dst(point, points_to_line(*l)), lines))
+
 size = (520, 390)
 
+def generate_models(sgrid, lh):
+    for f in [0, 1, 2, 3, 5, 7, 8, 11, 15, 17]:
+        grid = gm.fill(sgrid[0], sgrid[1], lh , f)
+        grid = [sgrid[0]] + grid + [sgrid[1]]
+        for s in xrange(17 - f):
+            grid = [gm.expand_left(grid, lh)] + grid
+        yield grid
+        for i in xrange(17 - f):
+            grid = grid[1:]
+            grid.append(gm.expand_right(grid, lh))
+            yield grid
+
+def score(grid, points, limit):
+    d = max(map(lambda l: dst((0, 0), points_to_line(*l)), grid + grid))
+    if d > limit:
+        return 0
+    return len([p for p in points if nearest2(grid, p) <= 2])
+
 points = pickle.load(open('edges.pickle'))
 
 lines = pickle.load(open('lines.pickle'))
@@ -40,36 +64,45 @@ r_lines = pickle.load(open('r_lines.pickle'))
 
 l1, l2 = lines
 
+lines_general = map(to_general, sum(lines, []))
+near_points = [p for p in points if nearest(lines_general, p) <= 2]
+
 while True:
-    l1s = random.sample(l2, 2)
+    t0 = time.time()
+    #l1s = random.sample(l1, 2)
+    l1s = [l1[0], l1[-1]]
     l1s.sort(key=lambda l: l[1])
-    corners = map(lambda l:linef.line_from_angl_dist(l, size), l1s) 
-    middle = ((0, 195),(520, 195))
-    # TODO! can I assume anything to be perspectively disorted square?
-    # TODO! take lower and middle and construct top
-    lh = (gm.intersection(corners[0], middle), gm.intersection(corners[1], middle))
-    grid = gm.fill(corners[0], corners[1], lh , 3)
-    grid = [corners[0]] + grid + [corners[1]]
-    grid.append(gm.expand(grid[-2], grid[-1], ((gm.intersection(middle, grid[-2]),
-                                         (gm.intersection(middle, grid[-1]))))))
-    grid.append(gm.expand(grid[-2], grid[-1], ((gm.intersection(middle, grid[-2]),
-                                         (gm.intersection(middle, grid[-1]))))))
+    sgrid = map(lambda l:linef.line_from_angl_dist(l, size), l1s) 
+    middle = lambda m: ((m, 0),(m, 390))
+    middle = middle(gm.intersection((sgrid[0][0], sgrid[1][1]), 
+                                    (sgrid[0][1], sgrid[1][0]))[0])
+    lh = (gm.intersection(sgrid[0], middle), gm.intersection(sgrid[1], middle))
+    sc, grid = max(map(lambda g: (score(g, points, 400), g), generate_models(sgrid, lh)))
     map(lambda l: pyplot.plot(*zip(*l), color='b'), grid)
-
-    plot_line(l1s[0], 'g')
+    #l2s = random.sample(l2, 2)
+    l2s = [l2[0], l2[-1]]
+    l2s.sort(key=lambda l: l[1])
+    sgrid = map(lambda l:linef.line_from_angl_dist(l, size), l2s) 
+    middle = lambda m: ((0, m),(520, m))
+    middle = middle(gm.intersection((sgrid[0][0], sgrid[1][1]), 
+                                    (sgrid[0][1], sgrid[1][0]))[1])
+    lh = (gm.intersection(sgrid[0], middle), gm.intersection(sgrid[1], middle))
+    sc, grid = max(map(lambda g: (score(g, points, 530), g), generate_models(sgrid, lh)))
+    print time.time() - t0
+
+    pyplot.scatter(*zip(*near_points))
+    map(lambda l: pyplot.plot(*zip(*l), color='b'), grid)
+    plot_line(l2s[0], 'r')
+    plot_line(l2s[1], 'r')
+    plot_line(l1s[0], 'r')
     plot_line(l1s[1], 'r')
-    
     pyplot.xlim(0, 520)
     pyplot.ylim(0, 390)
     pyplot.show()
 
 sys.exit()
 
-lines_general = map(to_general, sum(lines, []))
-
-near_points = [p for p in points if nearest(lines_general, p) <= 2]
 
-pyplot.scatter(*zip(*near_points))
 
 
 for l in lines[0]: