We propose a work based on the two classical methods, CUSUM and Shiryaev-Roberts, for the online change-point detection. These methods are based on sequential tests of likelihood ratio using recursive detection statistics. Our work is mainly built on the non-parametric versions of these approaches, suggested in the article [Tartakovsky, A. G. and all (2013)], by replacing the likelihood. We have proposed an extension of the detection procedures by using dynamic threshold detection which depends on time, and other rule-stops upon which detection procedures are based in the purpose of controlling especially the False Alarm Rate (FAR) and Average Delay of Detection (ADD). In order to assess the methods, we have performed an extensive simulation study in the purpose to exhibit the effect of the different parameters of the statistics and that of the stopping rules on the FAR and the ADD.