Traditional data integration solutions in the presence of the structure are not precise. During the integration period, the integrated result due to various reasons has many abnormal outliers, which cannot be detected and avoided with effective measures. This paper proposes an improved three stage outlier detection algorithm based on angle, which is mainly to detect the results after data integration, and effectively solve the problem of a large number of suspected outliers. This improved algorithm reduces the possibility of outliers in the traditional algorithm, taking into account the factors of dynamic changes in the data, verifying the abnormal real situation of suspected outliers for two times. This paper is backgrounded on the project of production accident emergency rescue system.