From ed2f10b265485040197402bbd5dbe0dd6058cb11 Mon Sep 17 00:00:00 2001
From: Tomas Musil <tomik.musil@gmail.com>
Date: Sun, 6 Jul 2014 00:25:52 +0200
Subject: [PATCH 1/1] preliminary work on generalized RANSAC

---
 src/gridf3.py | 18 ++++++++++++++++--
 src/imago.py  |  4 ++++
 src/ransac.py | 33 +++++++++++++++++++++++----------
 3 files changed, 43 insertions(+), 12 deletions(-)

diff --git a/src/gridf3.py b/src/gridf3.py
index 0697b92..693b8ee 100644
--- a/src/gridf3.py
+++ b/src/gridf3.py
@@ -12,6 +12,9 @@ from geometry import l2ad
 class GridFittingFailedError(Exception):
     pass
 
+class BadGenError(Exception):
+    pass
+
 def plot_line(line, c, size):
     points = linef.line_from_angl_dist(line, size)
     pyplot.plot(*zip(*points), color=c)
@@ -32,12 +35,16 @@ class Diagonal_model:
                 if l1[i] and l2[j]:
                     yield (l1[i], l2[j])
 
+    def remove(self, data):
+        self.data = list(set(self.data) - set(data))
+
     def initial(self):
         try:
-            return self.gen.next()
+            nxt = self.gen.next()
         except StopIteration:
             self.gen = self.initial_g()
-            return self.gen.next()
+            nxt = self.gen.next()
+        return nxt
 
     def get(self, sample):
         if len(sample) == 2:
@@ -50,11 +57,18 @@ class Diagonal_model:
         score = 0
         a, b, c = est
         dst = lambda (x, y): abs(a * x + b * y + c) / sqrt(a*a+b*b)
+        l1 = None
+        l2 = None
         for p in self.data:
             d = dst(p)
             if d <= dist:
                 cons.append(p)
+                if p.l1 == l1 or p.l2 == l2:
+                    return float("inf"), []
+                else:
+                    l1, l2 = p.l1, p.l2
             score += min(d, dist)
+
         return score, cons
 
 def intersection((a1, b1, c1), (a2, b2, c2)):
diff --git a/src/imago.py b/src/imago.py
index b513253..b87da99 100755
--- a/src/imago.py
+++ b/src/imago.py
@@ -108,6 +108,10 @@ def main():
                 d_file.close()
         else:
             lines, l1, l2, bounds, hough = linef.find_lines(image, do_something, logger)
+            #d_file = open('lines09.pickle', 'wb')
+            #pickle.dump(lines, d_file)
+            #d_file.close() #TODO delete this
+
 
         grid, lines = gridf.find(lines, image.size, l1, l2, bounds, hough,
                                  show_all, do_something, logger)
diff --git a/src/ransac.py b/src/ransac.py
index 4908421..cc16d2f 100644
--- a/src/ransac.py
+++ b/src/ransac.py
@@ -52,18 +52,20 @@ def iterate(model, distance):
     score = float("inf")
     consensual = model.initial()
     estimate = model.get(consensual)
-    new_score, consensual = model.score(estimate, distance)
-    while (new_score < score):
-        score = new_score
-        try:
-            estimate = model.get(consensual)
-            new_score, consensual = model.score(estimate, distance)
-        except NP.linalg.LinAlgError:
-            pass
+    new_score, new_consensual = model.score(estimate, distance)
+    if new_consensual != []:
+        while (new_score < score):
+            score, consensual = new_score, new_consensual
+            try:
+                estimate = model.get(consensual)
+                new_score, new_consensual = model.score(estimate, distance)
+            except (NP.linalg.LinAlgError):
+                pass
     return score, estimate, consensual
         
-def estimate(data, dist, k, modelClass=Linear_model):
-    model = modelClass(data)
+def estimate(data, dist, k, modelClass=Linear_model, model=None):
+    if not model:
+        model = modelClass(data)
     best = float("inf")
     estimate = None
     consensual = None
@@ -82,3 +84,14 @@ def ransac_duo(data, dist, k, mk, modelClass=Linear_model):
         model, cons = estimate(set(data) - set(cons), dist, k, modelClass)
     return (model, cons), estimate(set(data) - set(cons), dist, k, modelClass)
 
+def ransac_multi(m, data, dist, k, modelClass=Linear_model, model=None):
+    ests = []
+    cons = []
+    for i in xrange(m):
+        est, cons_new = estimate(None, dist, k, model=model)
+        model.remove(cons_new)
+        ests.append(est)
+    return ests
+
+        
+
-- 
2.4.2