public Tuple<string, int> Recognize(Image<Gray, byte> src)
{
var list = src.getEdge().Sample(100);
#region 计算形状上下文
int r = 15;
var scall = Jim.OCR.ShapeContext2D.ShapeContext.ComputeSC(list);
var scq = new double[r, 60];
for (int i = 0; i < r; ++i) {
int pos = rand.Next(list.Count);
for (int k = 0; k < 60; ++k) {
scq[i, k] = scall[pos, k];
}
}
#endregion
#region 计算距离
var dists = new ValueIndexPair<double>[SC.template_sc.Length];
for (int i = 0; i < SC.template_sc.Length; ++i) {
double[,] sci = SC.template_sc[i];
double[,] costmat = Jim.OCR.ShapeContext2D.ShapeContext.HistCost(scq, sci);
double cost = 0;
for (int j = 0; j < r; ++j) {
double min = double.MaxValue;
for (int u = 0; u < 100; ++u) {
double val = costmat[j, u];
if (val < min) min = val;
}
cost += min;
}
dists[i] = new ValueIndexPair<double> { Value = cost, Index = i };
}
#endregion
#region 对结果排序
var arr = dists
.OrderBy(p => p.Value)
.Select(p => new { Distance = p.Value, Char = SC.template_chars[p.Index] })
//.Where(p => "ABCDEFGHIJKLMNPQRSTUVWXYZ123456789".IndexOf(p.Char) != -1)
.Where(p => Filter.IndexOf(p.Char) != -1)
.ToArray();
Dictionary<string, int> matchcount = new Dictionary<string, int>();
foreach (var pair in arr.Take(Knn)) {
string ch = pair.Char;
matchcount[ch] = matchcount.ContainsKey(ch) ? matchcount[ch] + 1 : 1;
}
var match = matchcount.Select(pair => new { Count = pair.Value, Ch = pair.Key })
.Where(v => v.Count > 0)
.OrderByDescending(v => v.Count).ToArray();
//string result = "";
//foreach (var m in match.Take(3)) {
// result += String.Format("Char:'{0}',Accuracy:{1}/{2}\t", m.Ch, m.Count, knn);
//}
#endregion
return new Tuple<string, int> { First = match[0].Ch, Second = match[0].Count };
}
public Tuple<string, int> Recognize(Image<Gray, byte> challenge)
{
var list = challenge.getEdge().Sample(100);
var sc = ShapeContext.ComputeSC2(list);
if (sc.Length < 100) {
var tmp = new double[100][];
for (int i = 0; i < 100; ++i) tmp[i] = new double[60];
for (int i = 0; i < sc.Length; ++i) {
Array.Copy(sc[i], tmp[i], 60);
}
sc = tmp;
}
#region 量化到shapeme
int[] histogram = new int[100];
for (int i = 0; i < 100; ++i) {
double[] ds = new double[100];
for (int j = 0; j < 100; ++j)
ds[j] = ShapeContext.HistCost(sc[i], SC.shapemes[j]);
int id = ds.Select((v, idx) => new ValueIndexPair<double> { Value = v, Index = idx })
.OrderBy(p => p.Value)
.First().Index;
++histogram[id];
}
#endregion
#region 计算距离
double[] dists = new double[SC.template_histograms.Length];
for (int i = 0; i < SC.template_histograms.Length; ++i) {
dists[i] = Jim.OCR.ShapeContext2D.ShapeContext.ChiSquareDistance(histogram, SC.template_histograms[i]);
//dists[i] = Jim.OCR.ShapeContext2D.ShapeContext.HistCost(histogram.Cast<double>().ToArray(), templatehistograms[i].Cast<double>().ToArray());
}
#endregion
#region 对结果排序
var arr = dists.Select((d, i) => new ValueIndexPair<double> { Value = d, Index = i })
.OrderBy(p => p.Value)
.Select(p => new { Distance = p.Value, Char = SC.template_chars[p.Index] })
//.Where(p => "ABCDEFGHIJKLMNPQRSTUVWXYZ123456789".IndexOf(p.Char) != -1)
.Where(p => Filter.IndexOf(p.Char) != -1)
.ToArray();
Dictionary<string, int> matchcount = new Dictionary<string, int>();
foreach (var pair in arr.Take(Knn)) {
string ch = pair.Char;
matchcount[ch] = matchcount.ContainsKey(ch) ? matchcount[ch] + 1 : 1;
}
var match = matchcount.Select(pair => new { Count = pair.Value, Ch = pair.Key })
.Where(v => v.Count > 0)
.OrderByDescending(v => v.Count).ToArray();
//string result = "";
//foreach (var m in match.Take(3)) {
// result += String.Format("Char:'{0}',Accuracy:{1}/{2}\t", m.Ch, m.Count, Knn);
//}
#endregion
return new Tuple<string, int> { First = match[0].Ch, Second = match[0].Count };
}