Sentiment analysis in micro-blogging is an important task in mining social media, and has important theoretical and application value in personalized recommendation and public opinion analysis. Topic sentiment models have attracted much attention due to their good performance and ability of synchronized topic and the sentiment analysis in micro-blogs. However, most existing models simply assume that topic sentiment distributions of different micro-blogs are independent, which is contrary to the realistic status in micro-blogging and thus further leads to unsatisfactory modeling of micro-blogger's true sentiment. To address the issues, a probabilistic model, SRTSM (social relation topic sentiment model) is proposed. The new model introduces sentiment and micro-blogger social relation into LDA inference framework and achieves synchronized detection of sentiment and topic in micro-blogging. Extensive experiments on Sina Weibo show that SRTSM outperforms state-of-the-art unsupervised approaches including JST, SLDA and DPLDA significantly in terms of sentiment classification accuracy.