Calc registration

renderapps/registration/RansacAffineModel.py

@@ -0,0 +1,51 @@
+
+import ransac
+from numpy import vstack
+import cv2
+import numpy as np
+
+class RansacAffineModel(object):
+
+	def __init__(self,debug=False):
+		self.debug = debug
+
+	def fit(self,data):
+		data = data.T
+		#print("This is data shape")
+		#print(data)
+		fp = data[:2,:]
+		tp = data[2:,:]
+		fp = fp.T.astype(np.float32)
+		tp = tp.T.astype(np.float32)
+		return cv2.estimateRigidTransform(fp, tp,True)
+	def transformAffine(self, fp,A):
+		fp = fp.T
+		fp = np.vstack((fp,np.ones(fp.shape[1])))
+		newfp = np.dot(A,fp)
+		return newfp.T
+
+	def get_error(self,data,A):
+		if A is None:
+			A = np.array([[1.0, 0.0, 0.0],[0.0,1.0,0.0]])
+		data = data.T
+		fp = data[:2,:]
+		tp = data[2:,:]
+		fp = fp.T.astype(np.float32)
+		tp = tp.T.astype(np.float32)
+		#print(fp.shape)
+		#print(tp.shape)
+		#print(A)
+		#print(A.shape)
+		#print ("This is A: ")
+		#print(A)
+		#print("This is squrt")
+
+		fp_transformed = self.transformAffine(fp,A)
+
+		return np.sqrt(sum((tp.T-fp_transformed.T))**2)
+
+	def A_from_ransac(r,fp,tp,model,maxiter=5000,match_threshold=5):
+
+		data = vstack((fp,tp))
+		A,ransac_data = ransac.ransac(data.T,model,3,maxiter,match_threshold,0,debug=False,return_all=True)
+		return A,ransac_data['inliers']