An object-oriented multi-attribute differences matrix grey clustering method is proposed considering that the traditional grey clustering methods cannot make full use of the information relationships between the clustering objects and differences information of object’s multi-attributes. Grey class division is more general and the results of clustering have relatively strict constraints, lack of completeness. This method makes full use of existing multi-attribute information of the object, and can receive various, objective clustering results on different attribute category combination, making the classification of objects more realistic. Through the clustering analysis of state-level poverty-stricken counties in Anhui province, the results show the matrix grey clustering method’s application and indicates practicality and effectiveness of the matrix grey clustering method.