High-speed train gearbox shell is an important component of high-speed train. In order to protect the operational safety of high-speed train gearbox shell, it is needed to detect the casting internal defect as product testing and maintenance inspection accurately and rapidly. In this paper, based on three-dimensional CT technology the test was developed to detect the casting defects of high-speed train gearbox shell; through the analysis of three-dimensional data of the four kinds detects, three-dimensional geometric features and characteristic values were obtained, and the Adaboost_BTSVM algorithm were used to achieve the automatic classification of casting defects of high-speed train gearbox shell. The according classification accuracy of shrinkage defects can be 85%, and the classification accuracy of crack defects can stand at 100%. These will provide an available automatic identification method for the defect of high-speed train gearbox shell.