Machine vision-based vibrating screen motion characteristics rapid diagnostic method

Based on analysis of the mainstream vibrating screen motion characteristics diagnostic methods currently available, a rapid qualitative diagnostic method with the introduction of the Lucas-Kanade optical flow estimation algorithm and working based on characteristic point correspondence is proposed. The method can perform regular capture and analysis of motion characteristics of vibrating screen by using the light flow vector of feature points to describe the projection relationship of dynamic targets in 2 adjacent frames, detecting the number of feature points and light flow vectors in the target area in the continuous image sequence, setting the light flow threshold, and applying the positive and reverse error correction theory. Result of field test shows the method can effectively eliminate error tracking caused by system noise and shedding of fine coal particles; the algorithmic time for light flow tracking takes about 2.9 ms which is well within the predefined evaluation time of 15 s; and by using the lower-limit threshold evaluation method, the disorder in motion of vibrating screen caused by any local recessive mechanical fault can get rapidly diagnosed in a micro, accurate and qualitative manner.