A monocular SLAM forest test system was constructed, which used an ordinary smart phone embedded with an area array camera and an IMU as the hardware system, and used monocular SLAM technology to obtain depth maps, poses, etc. as data source. A method was designed based on smoothness and high robustness to filter the point cloud and tangent on the surface of the chest height cylinder; then based on the distance from the point to the surface of the cylinder and the distance from the tangent of the cylinder to the surface of the cylinder, an accurate estimation algorithm for the diameter at breast height and the standing tree position was constructed. Finally, based on the algorithm, an augmented reality tree measurement system was developed on the smart phone side, that was, real-time tree measurement by smartphones, and real-time manual supervision of the measurement results through augmented reality scenes. A highly robust method for filtering the point cloud and tangent on the surface of the breast height cylinder through smoothness was designed; then an accurate estimation algorithm for DBH and standing tree position was constructed based on the distance from the point and tangent to the surface of the cylinder. Finally, an augmented reality tree measurement system was developed on the smart phone end, which used the smartphone to measure trees in real time, and real-time manual supervision of the measurement results through the augmented reality scene. The tree measurement system was tested in five 32m×32m square plots to evaluate the usability of the tree measurement system at the same time; more importantly, each sample plot was investigated through single observation, orthogonal observation, symmetrical observation and surrounding observations in order to evaluate the impact of different observation methods on the accuracy of tree measurement. The results show that the deviation of the estimated standing tree position in the X and Y axes directions was -0.014~0.020m, and the root mean square error range was 0.04~0.08m; the deviation of the estimated DBH of standing tree was -0.85~-0.03cm (-3.60%~-0.04%), the root mean square error was 1.32~2.51cm (6.41%~12.33%). Compared with the single observation method, other observation methods can obtain higher accuracy estimation (especially for standing tree trunks that cannot be approximated as cylinders), orthogonal observation and symmetric observation were the best observation methods from the perspective of accuracy and efficiency. The results showed that the monocular SLAM augmented reality tree measurement system was a potential solution for accurate forest plot survey. 