A method is presented to reconstruct the three-dimensional (3D) canopy and stem based on terrestrial laser scanning
(TLS) data. Multiple range images are obtained and aligned on a 25 m x 25 m plot, where the single specie conifer forests
are growing in Gansu province, China. With regard of organizing the raw point cloud, firstly, a kind of kd-tree structure
is built to make the spatial index of them, and then, a kind of region grow is performed to segment them to smaller size,
and third step, an algorithm is make to extract a single canopy and stem. The segmentation processing is based on
"connected constraint", which can segment each individual tree in one subset. RANSAC and GAUSS image are used to
find cylinder in 3D point cloud to obtain the individual tree measurements, including the tree position and DBH. Each
cylinder is parameterized by its orientation and radius and is estimated iteratively.