Based on the daily closing price indices of Shanghai and Shenzhen stock market， this paper empirically tests structural break points of return volatility of China’s stock markets by using the modified ICSS algorithm，and two methods of dummy variables and standard deviation filtering are used to eliminate the influence of structural breaks respectively，then the GARCH and FIGARCH models are built to comparatively analyze the characteristics of stock market returns volatility before and after modification，and to mine the actual influence of structural breaks on stock market return volatility. The results find that the return volatility of Shanghai and Shenzhen stock market shows long-term memory and structural breaks lead to the overestimate of long-term memory of the return volatility，which reveals that China’s stock market does not reach weakly effective level. Financial quantitative model based on “effective market hypothesis” is not completely fitting for China’s stock market，the modified ICSS algorithm can effectively receive the structural breaks point of the volatility，virtual variables and standard deviation filtering can better get rid of the influence of structural breaks，however，the empirical result by using standard deviation filtering is relatively better. China’s stock return volatility has obviously structural breaks，and the time for structural breaks is all responding to important policies and market events，which indicates that China’s securities market is significantly affected by economic policies and market activities，thus，China should try to keep relative stability of the policies，reduce excessive administrative interference，boost marketization operation of the stock market，and closely concern about the possible attack of foreign and domestic economic development situation on China’s securities market.