track_keypoints() [2/2]

MAYAFLUX_API std::vector< TrackResult > MayaFlux::Kinesis::Vision::track_keypoints	(	std::span< const float >	prev_gray,
		std::span< const float >	curr_gray,
		uint32_t	w,
		uint32_t	h,
		std::span< const glm::vec2 >	prev_points,
		uint32_t	window_radius = 7,
		uint32_t	max_iterations = 20,
		float	eigen_threshold = 1e-4F,
		float	error_threshold = 0.3F
	)

Track keypoints from prev_gray to curr_gray via Lucas-Kanade.

For each input point, solves the Lucas-Kanade optical flow equation within a window_radius x window_radius patch using iterative Newton-Raphson refinement. Iteration stops when the displacement update falls below 0.03 pixels or max_iterations is reached.

A point is marked tracked=false if:

• The structure tensor within its window is near-singular (min eigenvalue below eigen_threshold)
• The tracked position moves outside the image boundary
• The residual error after convergence exceeds error_threshold

prev_points and the returned vector have the same length and order. Callers should discard points where tracked == false before passing to subsequent frames.

Parameters

prev_gray	Previous frame, single-channel float, size w * h.
curr_gray	Current frame, single-channel float, size w * h.
w	Image width in pixels.
h	Image height in pixels.
prev_points	Points to track in normalised [0, 1] coordinates.
window_radius	Half-size of the tracking patch in pixels.
max_iterations	Maximum Newton-Raphson iterations per point.
eigen_threshold	Minimum eigenvalue for structure tensor validity.
error_threshold	Maximum residual error to accept a track.

Returns

TrackResult per input point, same order.

Definition at line 171 of file OpticalFlow.cpp.

{
    auto grad = sobel(prev_gray, w, h);
    return track_keypoints(prev_gray, curr_gray,
        grad.dx, grad.dy,
        w, h, prev_points,
        window_radius, max_iterations,
        eigen_threshold, error_threshold);
}

References h, sobel(), and track_keypoints().

Referenced by MayaFlux::Kinesis::Vision::VisionExecutor::run(), and track_keypoints().

Here is the call graph for this function:

Here is the caller graph for this function: