Abstract: The calculation of mineral resources reserves is helpful to the rational development and utilization of mineral resources. At present,multi-point geostatistics is a popular technology for the calculation of mineral resources reserves,but it is difficult to construct high-quality three-dimensional training images in multi-point geostatistics. Therefore,a 3D training image construction method for mineral resource reserve calculation was proposed. Firstly,the 3D geological model is constructed based on the original borehole data,and the volume element model was obtained by using the irregular tetrahedral mesh. Secondly, based on the borehole attribute data,the regular volume element grade model was constructed by interpolation algorithm,and the attribute was assigned by Euclidean distance to the volume element model to obtain the 3D training image. Finally,the Snesim algorithm improved by multi-point geostatistics was used to construct a search tree to obtain the conditional probability,and the attribute of the volume element model was assigned based on a attribute fusion of correlative,and the mineral resources reserves were calculated. Taking the Heiniudong Ⅰ -3 orebody as an example,the average grade of copper in Ⅰ -3 orebody is calculated to be 2. 90%,the average grade of copper after mining is 2. 33%,the difference between them is only 0. 57 percentage points,which proves the correctness and effectiveness of the proposed method

Key words: multiple-point geostatistics,three-dimensional training image,attribute modeling of mineral body,calculation of mineral resources reserves