public override PointF MatchLocation(Fingerprint value, Match_Strategy strategy, double[] args)
{
switch(strategy)
{
case Match_Strategy.Single:
{
double min_dist = double.MaxValue;
PointF best_loc = new PointF();
foreach (KeyValuePair<PointF, Fingerprint> kvp in this.loc_fingerprint_dict)
{
double dist = (kvp.Value as Fingerprint_S).CalcSimilarity(value as Fingerprint_S);
if (dist < min_dist)
{
min_dist = dist;
best_loc = kvp.Key;
}
}
return best_loc;
}
case Match_Strategy.KNN:
{
if (args == null || args.Length < 1)
throw new Exception("KNN匹配方法需要额外一个参数");
Dictionary<PointF, double> dists = new Dictionary<PointF, double>();
foreach (KeyValuePair<PointF, Fingerprint> kvp in this.loc_fingerprint_dict)
{
double dist = (kvp.Value as Fingerprint_S).CalcSimilarity(value as Fingerprint_S);
dists[kvp.Key] = dist;
}
IOrderedEnumerable<KeyValuePair<PointF, double>> sorted = dists.OrderBy(x => x.Value);
int k = (int)args[0];
int i = 0;
KeyValuePair<PointF, double>[] NNs = new KeyValuePair<PointF, double>[k];
double sumr = 0;
foreach(KeyValuePair<PointF, double> kvp in sorted)
{
if (i >= k)
break;
NNs[i] = kvp;
sumr += 1 / kvp.Value;
i++;
}
double mean_x = 0;
double mean_y = 0;
foreach(KeyValuePair<PointF, double> kvp in NNs)
{
mean_x += kvp.Key.X * ((1 / kvp.Value) / sumr);
mean_y += kvp.Key.Y * ((1 / kvp.Value) / sumr);
}
return new PointF((float)mean_x, (float)mean_y);
}
}
return new PointF();
}
/// <summary>
/// 返回最匹配的位置坐标
/// </summary>
/// <param name="value"></param>
public PointF MatchLocation(Fingerprint value, Match_Strategy strategy, double[] args)
{
switch (strategy)
{
case Match_Strategy.Single:
{
double min_dist = double.PositiveInfinity;
PointF best_loc = new PointF();
foreach (KeyValuePair<PointF, Dictionary<Fingerprint_Name, Fingerprint>> kvp in this.loc_fingerprint_dict)
{
double dist = double.PositiveInfinity;
switch(value.Name)
{
case Fingerprint_Name.SA:
dist = (kvp.Value[value.Name] as Fingerprint_SA).CalcSimilarity(value as Fingerprint_SA);
break;
case Fingerprint_Name.S:
dist = (kvp.Value[value.Name] as Fingerprint_S).CalcSimilarity(value as Fingerprint_S);
break;
}
if (dist < min_dist)
{
min_dist = dist;
best_loc = kvp.Key;
}
}
return best_loc;
}
case Match_Strategy.KNN:
{
if (args == null || args.Length < 1)
throw new Exception("KNN匹配方法需要额外一个参数");
Dictionary<PointF, double> dists = new Dictionary<PointF, double>();
foreach (KeyValuePair<PointF, Dictionary<Fingerprint_Name, Fingerprint>> kvp in this.loc_fingerprint_dict)
{
double dist = double.PositiveInfinity;
switch(value.Name)
{
case Fingerprint_Name.SA:
dist = (kvp.Value[value.Name] as Fingerprint_SA).CalcSimilarity(value as Fingerprint_SA);
break;
case Fingerprint_Name.S:
dist = (kvp.Value[value.Name] as Fingerprint_S).CalcSimilarity(value as Fingerprint_S);
break;
}
dists[kvp.Key] = dist;
}
IOrderedEnumerable<KeyValuePair<PointF, double>> sorted = dists.OrderBy(x => x.Value);
int k = (int)args[0];
int i = 0;
KeyValuePair<PointF, double>[] NNs = new KeyValuePair<PointF, double>[k];
double sumr = 0;
foreach (KeyValuePair<PointF, double> kvp in sorted)
{
if (double.IsNaN(kvp.Value))
continue;
if (i >= k)
break;
NNs[i] = kvp;
sumr += 1 / kvp.Value;
i++;
}
double mean_x = 0;
double mean_y = 0;
foreach (KeyValuePair<PointF, double> kvp in NNs)
{
mean_x += kvp.Key.X * ((1 / kvp.Value) / sumr);
mean_y += kvp.Key.Y * ((1 / kvp.Value) / sumr);
}
return new PointF((float)mean_x, (float)mean_y);
}
case Match_Strategy.KNN_HC:
{
if (args == null || args.Length < 1)
throw new Exception("KNN匹配方法需要额外一个参数");
Dictionary<PointF, double> dists = new Dictionary<PointF, double>();
foreach (KeyValuePair<PointF, Dictionary<Fingerprint_Name, Fingerprint>> kvp in this.loc_fingerprint_dict)
{
double dist = double.PositiveInfinity;
switch (value.Name)
{
case Fingerprint_Name.SA:
dist = (kvp.Value[value.Name] as Fingerprint_SA).CalcSimilarity(value as Fingerprint_SA);
break;
case Fingerprint_Name.S:
dist = (kvp.Value[value.Name] as Fingerprint_S).CalcSimilarity(value as Fingerprint_S);
break;
}
dists[kvp.Key] = dist;
}
IOrderedEnumerable<KeyValuePair<PointF, double>> sorted = dists.OrderBy(x => x.Value);
int k = (int)args[0];
int i = 0;
List<MathUtils.MathUtility.Cluster> clusters = new List<MathUtils.MathUtility.Cluster>();
Dictionary<PointF, double> NNs = new Dictionary<PointF,double>();
foreach (KeyValuePair<PointF, double> kvp in sorted)
{
if (double.IsNaN(kvp.Value))
continue;
if (i >= k)
break;
NNs.Add(kvp.Key, kvp.Value);
clusters.Add(new MathUtils.MathUtility.Cluster(new List<PointF>() { kvp.Key }));
i++;
}
new MathUtils.MathUtility().HierachicalClustering_InPlace(clusters, 2);
MathUtils.MathUtility.Cluster cluster = clusters.OrderByDescending(x => x.PointNum).First();
double sum_r = 0;
double wx_sum = 0;
double wy_sum = 0;
foreach(PointF pt in cluster.point_list)
{
sum_r += 1 / NNs[pt];
wx_sum += pt.X * 1 / NNs[pt];
wy_sum += pt.Y * 1 / NNs[pt];
}
return new PointF((float)(wx_sum / sum_r), (float)(wy_sum / sum_r));
}
}
return new PointF();
}
