Point Cloud Segmentation


Point Cloud Segmentation can directly segment LiDAR point cloud, which can reduce the influence of under-canopy information loss in the CHM segmentation method. Individual tree information, including tree location, tree height, crown diameter, crown area and crown volume can be obtained from the segmentation results.


The point cloud segmentation algorithm was developed by Li et al.(2012). This method assumes that there are always gaps between trees. By finding the local maximum as seed points, each individual tree can be segmented based on the geometric correlations between each point and the seed points. The principle of this method is shown below:

Point Cloud Segmentation

The algorithm starts from the global maximum value, using it as the seed point A of the first segmented tree. Then, another local maximum, B, which is close to A but with a distance to A dAB larger than the user-defined threshold, will be used as the seed point of the second segmented tree. Every point between A and B will be clustered to one of these two trees following the minimum spacing rule. For example, if the spacing between point C and the first tree dAC is less than that between point C and the second tree dBC, point C will be clustered to the first tree correspondingly; if the spacing between point D and the first tree dAD is larger than that between point D and the second tree dBD, point D will be clustered to the second tree; if a point (point E) has the same distances to point A and point B, it will be set as the boundary point of these two trees. The spacing threshold should be close to the radius of the individual tree crown. When the spacing threshold is too large or too small, under-segmentation or over-segmentation may occur.


To identify individual trees using point cloud segmentation, navigate to and click on ALS Forest > Segmentation > Point Cloud Segmentation.

Point Cloud Segmentation


  • From Class: Classes which participate in the point cloud segmentation (All classes by default).
  • Input Data: Ensure that each input point cloud data is Normalize by DEM or Normalize by Ground Points. The input file can be a single file or multiple data files. Point cloud data should be opened in LiDAR360 before being segmented.
  • Spacing Threshold (meter)(default value is "2"): The 2D euclidean distance between the top of the tree is very important for tree segmentation. When setting the distance threshold parameter, the threshold value should be lower than the minimum allowable 2D Euclidean distance between two adjacent trees.
  • Height Above Ground (meter)(default value is "2"): Usually it is desirable to ignore points below a certain height to avoid the influence of low vegetation (e.g., grass and shrubs). Points below this threshold will not be considered in the segmentation. A value of 2 meters is commonly used.
  • Optimize color rendering for individual tree segmentation result (checked by default): By reorganizing the tree ID generated after the individual tree segmentation, it can greatly solve the problem of rendering the same color to the trees next to each other.
  • Output Path: Path of the output file. The resultant product is a comma-separated table in .csv format which contains the TreeID, location (x and y coordinates), height, crown diameter, crown area and crown volume of each individual tree. Refer to ALS point cloud segmentation results for the output example.
  • DefaultValue: Restore the default parameters.

Note: Only when the point cloud data are loaded in the software, can you use the Point Cloud Segmentation function; otherwise the message "There is no point cloud data meeting the conditions of calculation!" will pop up. If the maximum Z value of the point cloud is greater than 200 meters or the maximum Z minus the minimum Z is greater than 200 meters, the data is considered to not have been normalized, and the prompt information shown in the figure below will pop up. Click “YES” to keep using this type of data in the operation; otherwise, click "NO" and reselect the input data file.

Point Cloud Segmentation
        author={ Li W K, Guo Q H, Jakubowski M K and Kelly M},
        title={A new method for segmentation individual trees from the LiDAR point cloud},
        booktitle={ Photogrammetric Engineering and Remote Sensing,78(1):75-84},

results matching ""

    No results matching ""