In order to automatically and accurately extract the diameter at breast height (DBH) of trees within a certain range from the single-station terrestrial laser scanning（TLS）, an adaptive circle-ellipse fitting DBH estimation method is proposed based on the point cloud slice. The forestry point cloud data at breast height is sliced, the point cloud of the slice is clustered, and the proposed method is used to distinguish whether they are trunk points. DBH of the trunk point is calculated directly when the point conforms to the circular distribution, while DBH of the trunk point is calculated after breast height position correction when the point conforms to the elliptic distribution. The artificial willows data acquired in Beijing Dongsheng Country Park is used to verify the proposed method and compare it with the simple circular fitting method. The results show that the root mean square error of estimated DBH is 1.1 cm when the scanning distance is within 26 m, and it is 1.99 cm when the scanning distance is within 56 m. For the trees regarded as elliptic distribution, the mean error of DBH estimation results is 4.7% lower than that of the circular fitting results. The DBH can be detected quickly and efficiently with the proposed method.