Markerless froth velocity measurement in coal flotation via local motion-consistent ORB matching

The froth velocity in coal flotation is a key indicator reflecting the separation performance. However, full-foreground froth images often exhibit significant local similarity or suffer from motion blur, resulting in weak texture and poor velocity extraction accuracy. Existing feature-based methods typically rely on local intensity feature (LIF), which are insufficient for effective matching. Therefore, an enhanced ORB (Oriented FAST and Rotated BRIEF) descriptor with local motion consistency constraints is proposed. It integrates LIF with optical flow relative magnitude and direction (OF-RMD). Initial LIF is extracted using ORB, and the OF-RMD of each key point is calculated within a constrained window. LIF and OF-RMD are then weighted combined to guide point matching. Experiments show that the proposed method achieves an average matching recall of 87.36%, improving by 12.81% over standard ORB, with a velocity extraction accuracy of 84.07% and an average processing time of 25.41 ms per frame. This provides an interpretable and efficient solution for markerless froth motion measurement.