This invention relates to sintered mineral phase analysis techniques. It discloses a automatic identification method of major mineral phase in sintered minerals, the method is applicable for automatically find out the optimal threshold value under the condition that distribution of shade of gray of all types of sintered phase are mixed and the regular of peak value distribution is not obvious, that is based on the expectation maximization (EM) methods; using mathematical expectation maximization histogram approximation algorithm, through the fit of Gaussian distribution of different components to obtain the most optimized image segmentation gray threshold; analysis the digital microscopy image of the sintered minerals to acquire optimal threshold value parameters of segmentation of by hematite, magnetite, iron calcium, and silicate binder these five categories, so as to achieve the computer intelligent identification method of hematite, magnetite, iron calcium, and silicate binder, thereby achieving accurate statistical calculation of each phase content by computers, laid a solid and accurate basis for phase analysis of sintered mineral.