Warp MT::Lucas_Kanade(Warp warp)
{
for (int iter = 0; iter < max_iteration; ++iter) {
Matx61f G;
Matx<float, 6, 6> H;
G = 0.0f;
H = 0.0f;
float E = 0.0f;
for (int i = 0; i < fine_samples.size(); ++i) {
Matx<float, L4, 1> T(fine_model.ptr<float>(i)), F;
Matx<float, 2, L4> dF;
Matx<float, 2, 6> dW = warp.gradient(fine_samples[i]);
Point2f p = warp.transform2(fine_samples[i]);
feature.gradient4(p.x, p.y, F.val, dF.val, dF.val + L4);
T -= F;
float e = sigmoid(T.dot(T));
E += e;
float w = sigmoid_factor * e * (1.0f - e);
G += w * (dW.t() * (dF * T));
H += w * (dW.t() * (dF * dF.t()) * dW);
}
E = E / fine_samples.size();
Matx61f D;
solve(H, G, D, DECOMP_SVD);
warp.steepest(D);
if (iter > 1 && D(3) * D(3) + D(4) * D(4) + D(5) * D(5) < translate_eps) {
if (log != NULL)
(*log) << "\terror in iteration " << iter << " = " << E << endl;
break;
}
}
return warp;
}
float MT::evaluate(Warp warp)
{
rect_t rect = window(warp.t);
float fine_cell = sqrt(rectArea(rect) / cell_n);
feature.set_cell(fine_cell);
feature.set_step(1);
float E = 0.0f;
for (int i = 0; i < fine_samples.size(); ++i) {
Vector2f p = warp.transform2(fine_samples[i]);
Vector32f F;
feature.descriptor4(p.x(), p.y(), F.data());
E = E + sigmoid((F - fine_model.col(i)).squaredNorm());
}
return E / fine_samples.size();
}
float MT::evaluate(Warp warp)
{
Rect2f rect = window(warp.t);
float fine_cell = sqrt(rect.area() / cell_n);
feature.set_cell(fine_cell);
feature.set_step(1);
float E = 0.0f;
for (int i = 0; i < fine_samples.size(); ++i) {
Matx<float, L4, 1> T(fine_model.ptr<float>(i)), I;
Point2f p = warp.transform2(fine_samples[i]);
feature.descriptor4(p.x, p.y, I.val);
T -= I;
E = E + sigmoid(T.dot(T));
}
return E / fine_samples.size();
}
void MT::fine_train(Warp warp)
{
Rect2f rect = window(warp.t);
float fine_cell = sqrt(rect.area() / cell_n);
feature.set_cell(fine_cell);
feature.set_step(1);
Mat model(fine_samples.size(), L4, CV_32FC1);
for (int i = 0; i < fine_samples.size(); ++i) {
Point2f p = warp.transform2(fine_samples[i]);
feature.descriptor4(p.x, p.y, model.ptr<float>(i));
}
if (fine_model.empty()) {
N = 1;
fine_model = model;
}
else {
++N;
fine_model = (float(N - 1) / N) * fine_model + (1.0f / N) * model;
}
}
void MT::fine_train(Warp warp)
{
rect_t rect = window(warp.t);
float fine_cell = sqrt(rectArea(rect) / cell_n);
feature.set_cell(fine_cell);
feature.set_step(1);
//compute the features on the grid
MatrixXf model(32, fine_samples.size());
for (int i = 0; i < fine_samples.size(); ++i) {
Vector2f p = warp.transform2(fine_samples[i]);
feature.descriptor4(p.x(), p.y(), model.col(i).data());
}
//update using a moving averaging manner
if (N == 0) {
N = 1;
fine_model = model;
}
else {
++N;
fine_model = (float(N - 1) / N) * fine_model + (1.0f / N) * model;
}
}
