private void ransac_btn_Click(object sender, EventArgs e)
{
ransac = ApplyRansac();
//we need to sort the points, otherwise we will get a random figure
List <Point> sorted1 = new List <Point>();
List <Point> sorted2 = new List <Point>();
for (int j = 0; j < Shape1.Count; j++)
{
if (ransac.best_consensus1.Contains(Shape1[j]))
{
sorted1.Add(Shape1[j]);
sorted2.Add(Shape2[j]);
}
}
List <Point> new_Shape2 = ApplyTransformation(ransac.best_model, sorted2);
Pen pBlue = new Pen(Brushes.Blue, 1);
Pen pRed = new Pen(Brushes.Red, 1);
Graphics g = ransac_panel.CreateGraphics();
DisplayShape(sorted1, pBlue, g);
DisplayShape(new_Shape2, pRed, g);
MessageBox.Show("Cost = " + ransac.best_error);
}
Ransac_Algo ApplyRansac()
{
Ransac_Algo rans = new Ransac_Algo();
int k = 20; //iterations for the algorithm
int t = 7; //treshold to determine if we accept the new point or not, if it adds less than 7 to the cost it is valid
int n = 3; //number of initial random point
int d = 4; //minimum data points to accept the model as valid
int iterations = 0; //counter of iterations done
/*
* Transformation best_model = new Transformation();
* List<Point> best_consensus_set1 = new List<Point>();
* List<Point> best_consensus_set2 = new List<Point>();
* Ransac_Algo rans = new Ransac_Algo();
*/
double best_error = int.MaxValue;
Random rand = new Random(); //used to select random indexes
while (iterations < k)
{
List <int> random_index = new List <int>(); //list that will contain random indexes from shapes
List <Point> maybe_inliners1 = new List <Point>();
List <Point> maybe_inliners2 = new List <Point>();
Transformation maybe_model = new Transformation();
double maybe_cost = 0; //cost from the random points
//selecting random points (using indexes) from data
for (int i = 0; i < n; i++)
{
int index = rand.Next(Shape1.Count);
while (random_index.Contains(index))
{
index = rand.Next(Shape1.Count);
}
random_index.Add(index);
maybe_inliners1.Add(Shape1[index]);
maybe_inliners2.Add(Shape2[index]);
}
maybe_model = ICPTransformation.ComputeTransformation(maybe_inliners1, maybe_inliners2); //computing model for random chosen points
maybe_cost = ICPTransformation.ComputeCost(maybe_inliners1, maybe_inliners2, maybe_model); //computing cost for random chosen points
List <Point> consensus_set1 = new List <Point>(maybe_inliners1);
List <Point> consensus_set2 = new List <Point>(maybe_inliners2);
for (int i = 0; i < Shape1.Count; i++)
{
if (!random_index.Contains(i))
{
maybe_inliners1.Add(Shape1[i]);
maybe_inliners2.Add(Shape2[i]);
consensus_set1.Add(Shape1[i]);
consensus_set2.Add(Shape2[i]);
Transformation temp_model = ICPTransformation.ComputeTransformation(consensus_set1, consensus_set2);
double temp_cost = ICPTransformation.ComputeCost(consensus_set1, consensus_set2, temp_model);
double dif = Math.Abs(maybe_cost - temp_cost);
if (dif >= t)
{
consensus_set1.RemoveAt(consensus_set1.Count - 1);
consensus_set2.RemoveAt(consensus_set2.Count - 1);
}
}
}
if (consensus_set1.Count > d) //checking if consensus has enough points to make the model valid
{
Transformation better_model = ICPTransformation.ComputeTransformation(consensus_set1, consensus_set2);
double this_error = ICPTransformation.ComputeCost(consensus_set1, consensus_set2, better_model);
if (this_error < best_error)
{
best_error = this_error;
rans.best_model = better_model;
rans.best_consensus1 = consensus_set1;
rans.best_consensus2 = consensus_set2;
rans.best_error = this_error;
}
}
iterations++;
}
return(rans);
}