Find the object's motion vector - opencv

Question

JavaCV ~ OpenCV
It is necessary to determine the direction of a person's movement (left or right) in real time. I use JavaCV. My code gives the correct answer only based on movement in ideal conditions (there are no shadows, the lighting does not change, the person does not change the trajectory of movement):
We place the found moving object in a green rectangle, calculate the sum of the centers of mass of all rectangles, calculate the difference between the centers of mass of the previous and current frames, keep statistics from a few comparisons (to reduce the probability of an error in the final result) and on its basis we make a verdict - moves to the right or to the left.
I found the function calcGlobalOrientation(), but I don't have an example of how to use it.
Perhaps someone will give advice on how to use JavaCV with a fairly high performance and low probability of error to determine the direction of movement of the object?
if (cnts.size() > 0) {
firstEntry = true;
center_current = newPoint();
for (int i = 0; i < cnts.size(); i++) {
//continue to work only with areas that satisfy the condition
if (contourArea(cnts.get(i)) < 2000) {
continue;
}
Rect r = boundingRect(cnts.get(i));
// check if the center of mass of the figure is in the "gray" area
if (
isInside(r.x() + r.width() / 2, r.y() + r.height() / 2, getArrayGrayZone().get(0)) ||
isInside(r.x() + r.width() / 2, r.y() + r.height() / 2, getArrayGrayZone().get(1))
) {
// System.out.println("Movement in the \"grey\" zone");
continue;
}
//4th parameter - characterization of the contour of the figure ("< 0" the figure is filled, "> 0" the line thickness)
rectangle(frame, r, new Scalar(0, 255, 0, 0), 3, 0, 0);
if (firstEntry) {
//+1 frame with recognized motion that satisfies the condition
counterFrame++;
firstEntry = false;
}
//look for the center of mass of the given rectangle
center = newPoint();
center.x = r.x() + r.width() / 2;
center.y = r.y() + r.height() / 2;
//if the current center of mass has not yet been calculated, then enter a new value
if (center_current.x == 0) {
center_current = center;
continue;
}
//calculate the common center of mass of the current and past rectangles
center_current.x = (center_current.x + center.x) / 2;
center_current.y = (center_current.y + center.y) / 2;
}
//if it was possible to calculate the current center of mass of the rectangles
if (center_current.x != 0) {
//if the previous center of mass has not yet been calculated, then equalize the current and previous
if (center_prev.x == 0) {
center_prev = center_current;
} else {
difference = new Vec2d(center_current.x - center_prev.x, center_current.y - center_prev.y);
if (difference.x < 0) {
indexL++;
}
if (difference.x > 0) {
indexR++;
}
}
}
}
if (counterFrame % 10 == 0 && counterFrame > 0) {
if (indexR > indexL) {
System.out.println(
"Right" + df.format(new Date()));
}
if (indexL > indexR) {
System.out.println(
"Left" + df.format(new Date()));
}
//reset the counter of frames with a recognized motion that satisfies the condition
counterFrame = 0;
//reset the "statistical" variables for detecting shift to the right or left
indexL = indexR = 0;
}
//make current center of mass "previous"
center_prev = center_current;