In different terrain conditions, different feature combinations and dimensions have different influences on the effectiveness and accuracy of classification. A method is proposed to select the airborne LiDAR point cloud classification with adaptively feature selection. The whole point cloud is divided into different regions in accordance with the terrain conditions, and the suitable feature set is selected adaptively for classification. In order to evaluate the effective of this method, the random forest method and support vector machine classification method are used to classify the experimental data with the feature set after optimization. Experimental result shows that the suitable feature set for classification in different areas are different. The proposed method can reduce the feature dimensions effectively, shorten time consumption, and achieve high classification accuracy.