Aiming at the problem of intelligent defect detection for the circlular knitting machine in real-time production in China, according to the method of image quantitative analysis, a rotary cycle prediction method based on recursive least squares (RLS) adaptive filter is put forward. The background and the value of industrial application of using RLS algorithm to cycle prediction are firstly introduced. Secondly, the RLS algorithm and the basic working principle of RLS filter of using cycle prediction are analyzed in detail. Finally, in the real-time production of two-sided machine, the real cycle data needed for the prediction is obtained according to the up-down system of circular knitting machine. The real cycle is dealt by using the RLS filter to find its optimal order. On the basis of the best order, contrast experiments of cycle prediction of RLS adaptive filter and one-step cycle prediction are designed to prove that the RLS filter for circular knitting machine rotating cycle stability is superior to the one-step prediction method. Combining the different advantages of the two methods, the cycle prediction processing mechanism of an intelligent defect detection system is obtained, which is applied to the real-time control of industry site.