Abstract: A tangent plane segmentation algorithm of extracting stem point from tree point cloud which scanned by terrestrial laser scanning was presented. Firstly, based on the slicing and angular partition of the tree point cloud, the start slice was selected from the successive slices with little distortion under the branches. Next, the tangent plane of an angular partition was constructed according to the points that belong to the same, the prior and the next angular partition number of the adjacent segmented slices. The stem points of the angular partitions were extracted by the position relationship between the tree points and the tangent plane. Finally, the experiments were carried out on 13 poplar trees point scanned by terrestrial laser scanner when leaves fallen. The stem point was efficiently extracted, at the same time the characteristics of the stem surface were retained. Therefore the presented algorithm can provide base data for subsequent research of stem.