One of the main sectors in the weaving is the preparation, in which a lot of goals are pursued to improve the condition of weaving process. Principally winding is one stage of the weaving preparation that transfer the yarn of ring bobbin on the main bobbin. It is important to know that random winding , a kind of bobbin winding, has many benefits and applications in the textile industry. However this process causes a winding problems. In some of the diagonal packages, successive coils of yarns coincide with one another, and this effect persists till the package diameter increases sufficiently as further yarn is wound. The overwound turns of yarn tend to form a hard band, and hence the problem is termed patterning. The concept may cause many problems during winding and unwinding. Therefore recognizing the time of patterning before occurrence is an important concept to prevent the following mentioned problems. In this study, the time of patterning in the random yarn winding, is evaluated by image processing through the linear mean and variance. For this purpose, photographs of the bobbin surface were continuously taken with specified time interval. Then a two dimentional matrix was obtained for any picture by digitization the pictures. Next the mean and variance of any matrix was calculated. At the end, a diagram of means and variances were plotted, and a peak was perceived in place of patterning defect. By this peak the moment of occurrence of defect was diagnosed, and the message of “patterning” will be announced