Abstract: On the basis of characteristics of wind-sand two-phase flow CCDimages, an auto generated Digital Mask algorithm based on K-mean valuealgorithm of pattern recognition's dynamic clustering method is proposed tosolve the wind-sand two-phase flow. This algorithm simplifies the manualsetting operation of parameters in the process of the traditional generation ofDigital Mask, reduces the man-made error, and provides a safe and fastmethod to process two-phase PIV images in batch. Finally, this algorithm isapplied to real measurements of wind-sand two-phase flow. After computation,the flow fields of the phase of the wind, the phase of the sand and thetwo-phase of the wind-sand of different flow types are obtained.