In order to deal with the problem of the slow on-line testing speed which is caused by the traditional regression model that is not sparse enough,a method based on the relevance vector machine (RVM) regression is proposed in the inertial confinement fusion (ICF) experiment target gesture estimation. In the experiment,the algebra features extracted by the principal component analysis (PCA) is used as the sample features of the RVM,which can deal with the bluer problem caused by the small depth of field. A comparison between the commonly used regression algorithms such as support vector regression (SVR),K-nearest neighbor (KNN) and least mean square (LMS) has been done in this paper,the result shows that the testing error square of the RVM and SVM are the smallest,and the testing accuracy of them are the highest also. The testing time of the RVM is the lowest in the several algorithms,which shows that the RVM is more suitable for on-line testing.