分析了Horn-Schunck全局光流算法和Lucas-Kanade局域光流算法在条纹位移测量中的分辨力和测量范围,结果表明:当Horn-Schunck算法的相对误差和Lucas-Kanade算法的相对误差小于2%时,两种算法的相位分辨力都能够达到10 -13π,对应像面上的位移分辨力为1.6×10 -12 pixel,两种算法在理论上与四步相移法的分辨力相当;在有噪声的情况下,两种算法的分辨力都达到了0.01π,对应像面上的位移分辨力为0.16 pixel;在相对误差小于2%、方均根误差小于3%时,Horn-Schunck算法和Lucas-Kanade算法的测量范围分别为0~17π/100和0~52π/100,分别约为0~π/6和0~π/2,并且测量范围受噪声的影响很小。

In this work, the resolution and measuring range of the Horn-Schunck global optical flow algorithm and Lucas-Kanade local optical flow algorithm in fringe displacement measurement are analyzed. The simulation results show that when the relative error of the Horn-Schunck algorithm and Lucas-Kanade algorithm is less than 2%, and the resolution of Horn-Schunck algorithm and Lucas-Kanade algorithm can reach 10 ^-13π. Furthermore, the corresponding displacement resolution on image plane is 1.6×10 ^-12 pixel. This shows that the resolution of the two algorithms is consistent with that of the four-step phase shift method. In the case of noise, the resolution of both algorithms reaches 0.01π, and the displacement resolution of the corresponding image surface is 0.16 pixel. When the relative error is less than 2% and the root-mean-square error is less than 3%, the measurement ranges of Horn-Schunck algorithm and Lucas-Kanade algorithm are 0--17π/100 and 0--52π/100, approximately 0--π/6 and 0--π/2, respectively, and the measurement ranges are less affected by noise.